The title sounds to me like: I am going to spend $1000 in groceries and dance lessons. That is, two very different things lumped together.
Memory chips are like groceries, essential commodity parts, a no-nonsense investment. Humanoid robots are like dance lessons, it is cool, it is sexy, and it may pay off in the future, but the value is much less certain.
I don't have the historic numbers at hand, but I would assume that for each of those categories this is a similar proportional increase, so it's similarly notable to mention the increase in spending on humanoid robots.
Androids (humanoid robots) will require loads of ram, and loads of model training under the current paradigm. So it sort of makes sense. At least, I see robots as the top of the pyramid.
Autonomous (non-teleoperated) humanoid robots that can do useful work in an unfamiliar environment do not exist. And nobody's close enough to making them to understand if they're possible with our current level of technology, let alone how.
Most initial work for them would be in familiar, well-controlled environments - replacing humans in existing factories. I think whether they'd be cost effective for that will remain unknown even after a few years in service though.
Factory work that can be straightforwardly automated by robots generally already is, by special-purpose factory robots. The remaining tasks are generally:
* Tasks where poorly-paid humans are cheaper than expensive factory robots. Humanoid robots are more complex / fiddly so will be more expensive than existing factory robots. No help here.
* Tasks where human dexterity has an advantage over state-of-the-art robot actuators (e.g. sewing fabric panels into garments). Better robotics could help here, but the advancement needed is better actuators, not AI and a humanoid form-factor. And if you solve this you'd be better off putting your new end-effector on an existing 6dof platform.
* Supervising robotic equipment and handling exceptions. But then you get to "handling poorly-specified unfamiliar tasks in the physical world" which is not currently a solved problem and there's no guarantee just throwing more compute at it will be sufficient to solve this. So far all humanoid robot demos have either been either known tasks in a tightly-controlled environment, teleportation, or so poorly-functioning as to be obviously not fit for purpose.
I'd also point out that what makes this whole thing smell of being a grift is the fact that what's being chased is humanoid.
Humans are not the pinnacle of dexterity or stability. And freed from biological constraints, it makes no sense why we'd use the human form as a reference.
Making a humanoid robot is a hard thing to do, for sure, but it's also not particularly useful. The routines needed to balance a robot or correct for a slip are interesting to solve, but don't really make for a better robot which is more capable of doing the dishes or folding the laundry.
If I'm amazon, for example, then the most useful form factor for a general purpose robot is 2 arms on a 4 wheel omnidirectional rolling platform.
Our hands are actually amazing, to a degree that’s hard to comprehend unless you’ve tried to build something equivalent. Our finger tips can discern features down to around 10 microns, detect a wide range of vibrations and single shocks, manipulate objects two orders of magnitude smaller than them, grasp an incredible range of weird shapes and materials reliably, repeat pre-learned motions remarkably quickly, and with training carry kilograms, all while weighing about a kilogram and reliably operating maintenance free for tens of thousands of hours. Oh and they’re made of meat.
An artificial manipulator with even half of these features would be the holy grail of robotics. Doing them all at once is a miracle.
Completely agree. My point is more that even though our hands are amazing and even if we decided we want all the features of hands, there's no reason why you'd design such a device with 5 fingers and an oddly placed opposable thumb. There's no reason you'd have more than 2 joints per finger.
That's what I'm calling a grift of the humanoid robots. They are focusing pretty heavily on making manipulators shaped like hands, which don't perform anywhere near hand capabilities. The form is mattering more than the functionality which introduces a lot of unnecessarily hard to solve functionality problems.
The actual hard part of doing something like folding cloths isn't even the manipulation (though it's part of it) you mostly just need a few pincers. The hard part is that fabric changes like crazy in a 3 dimensional way on every motion. Just identifying "this is a shirt" is a hard problem to solve. Further deciding how to fold said shirt is extra difficult. But even further, getting a robot to know "this is where this shirt should be put away" and "this is how I should handle new cloths" is crazy hard. That's the part that I've seen absolutely no evidence that any of these humanoid robots are making any sort of progress on. The most impressive ones are cheating with a guy in the room over doing the task. Not exactly something I'd want in my home. Perhaps in a nuclear reactor.
I would say that turtles are the pinnacle of stability, and longevity, and so we have Roombas. But, Racoons are at the pinnacle of dexterity, and Octopodes are at the pinnacle of dimensional manipulation...
but personally, I would welcome our cyper-coon overlords.
So far all humanoid robot demos have either been either known tasks in a tightly-controlled environment, teleportation, or so poorly-functioning as to be obviously not fit for purpose. It's possible throwing more compute at the problem will work and they'll be useful in a few years. Or maybe we're actually experiencing a Markov-chain / ELIZA moment and are multiple decades away from anything actually useful.
Self driving cars is far from a solved problem. It's something we've been working on for the last 20 years at least. We are getting closer, some solutions are impressive (like waymo). But even those ultimately need operators in the area of the cars to solve the problem of the car getting stuck.
Self driving cars are an infinitely simpler problem to solve vs a general purpose humanoid robot. You have basically 2 outputs, acceleration and steering. You have rules simple enough that I was driving at 14. With enough input, you'd think self driving could be completed in a snap. But it's not there yet.
Humanoid robots which are useful will come after widespread deployment of L5 self driving cars. Since we don't have that, I have no faith that we are close to useful humanoids.
"I have no faith that we are close to useful humanoids." I agree completely.
We still have competitions about having robots walk up stairs and open doors,
and balance while doing those things...
To replace humans for general tasks. McDonalds would like to use them to flip burgers, take orders, and deliver meals. Amazon would like to use them to pick items from their warehouses for delivery.
I'd argue, though, in both cases the "humanoid" part isn't what they really want or need. Even in the mcdonalds case, so long as the robot is "cute" enough to interact with customers, then it can be a box with an arm.
Of course, a major problem with using robots in the food industry is cleaning and hygiene. Robots tend to have lots of hard to clean cracks that will happily carry around germs. Even in medicine, we are finding that things like laparoscopic devices are beasts that are almost impossible to fully clean.
I have been following them since they were Halodi robotics. They're cool, but nowhere near the level of autonomy you need. My theory is that before household robots become a thing we should have self driving cars be a common occurrence, since that is a much much simpler problem.
We said the same thing about Waymo, that it was perpetually in the future. It took them less than a decade. The robots today are functionally capable, they don’t have the right fuzzy intelligence yet. It’s purely a data problem (lack of) and a lot of people are working on it.
From what I've seen with Waymo, autonomous driving in relatively good conditions (light to medium rain, fog, or sunny) is a solved technical problem, even in small, winding streets. Snow is the next big hurdle, and they're actively working on that in Denver and Detroit.
They're being conservative with their rollout mostly because of civic issues. Most places do not have a legal framework for "what if your autonomous vehicle hits someone?" yet. Even if Waymos never were at fault for a collision with a person, you can always have cases like the one in Georgia back in October where a bicyclist wasn't looking where they were going and rammed into one. The shaky legal ground is a pretty big impediment right now, and that's in the US where we have much laxer laws about corporations killing people.
Yeah, I suspected that, but legislation and dealing with who is responsible and what happens if an autonomous vehicles directly kills/hurts someone or is involved in an accident it IS part of the problem, luckily.
Then, there will be concerns about the scaling costs. If it makes economical sense only in densely populated zones, what's the point? I mean, yes, it will be yet another business that works only in cities in that case, but that will not then make it universal as cars are.
If they can pull off year round driving in Manhattan, and they have, it’s a solved problem. Rollout it slow due to bureaucracy, appetite for progressiveness and catering to entrenched business interests but it will show up everywhere it’s profitable eventually.
It's not just a data problem, it's a hardware problem. Transformer-based robots require even more processing power than plain LLMs, as they also need to process visual and spatial/touch input. We don't have GPUs capable of fast inference on a SOTA LLM that would fit in a robot brain form factor, let alone also run fast enough spatial and visual processing. And there's currently nothing even approaching a feasible solution for cooling such a device.
If there's no unknown unknowns in the brain, it's most likely possible. As the universal approximation theorem and empirical results of scaling SGD+RL suggest. Whether it will be economically viable remains to be seen. The human cerebellum has a peculiar structure and 80% of the brain's neurons after all.
The parameter count equivalent of a human brain is not yet known, but if it was one per synapse then a full human brain replica would need about 1.5e14.
We also don't yet know how to be as efficient with training examples as any living creatures' brain, and we only partially make up for this by training on so many examples it would take you a million or so years to do the same, so we'd still stuggle with something proportionally smaller-brained such as a cat.
That said, remote controlled androids are going to be economically disruptive, as they make every (unlicensed) job open to outsourcing from an office in a low wage country.
Real neurons are orders of magnitude more complex than their artificial pseudo-approximation (it is all based on the century-old understanding of how neurons work). You can think of _individual_ biological neuron as an analog of the small artificial neural network. You can see this simple visual explanation on YouTube[1]. So we aren't even close. It doesn't mean the AI is impossible, it just means people underestimate the "computing power" of real brains, as well as that AI, even the future one might be totally different in how it works from the natural intelligence.
Brain's information flow also isn't just through direct neuron-to-neuron connections. Firing also releases neuromodulators into the extracellular space which affects how other neurons operate. Furhermore neuron connection architecture is very different between brain and feedforward ANNs, with the former exhibiting a lot of recurrent connections.
But for biological neurons to do something that can't be efficiently approximated on a digital computer (but conductive to useful information processing) they need to have unknown unknowns (well, partial unknowns like an unknown quantum algorithm will do too).
We don't know the violations of the physical Church-Turing thesis that are conductive for machine learning. We don't have evidence for their existence in the brain (although, the brain would be the prime candidate for finding them as evolution works directly with the true physical laws).
BTW, large ANNs don't try to model how the brain does things. They are trying to mimic what the brain does. So, using "how many transistors/artificial neurons it takes to model a biological neuron" is not a good approach.
We have no evidence. We even have no solid theories how this can work (Penrose's OrchOR is "OrchOR somehow taps into mathematical knowledge somehow encoded into the structure of spacetime"). But people, for some reason, insist that there should be something there. I can't attribute it to anything else but to deeply entrenched feeling of human exceptionalism.
You're talking about a philosophical debate whether the brain is computable, the other commenters are pointing out that even conservative estimates point to a brain-like NN requiring over a quadrillion parameters.
Let's define the terms. What does it mean "with similar capacity"? As far as I understand xvilka was taking about the number of artificial neurons required to model a biological neuron times the number of biological neurons in the human brain. It is modelling its physical structure (on a neuronal level).
That is we have unknowns that might increase or decrease estimates of computational demands of a functionally equivalent ANN. Not everything that happens in the brain contributes to information processing.
If you're running a massive model for logic you're probably better off not putting it in the robot. And it'll be a long time before there's enough robots to make up a significant share of usage.
More basic movement control doesn't need loads of ram as far as I know.
The same logic for why self-driving cars can't be cloud based, applies for robots. Something cannot be in the middle of a delicate operation and then "oops!", no network, it just stops.
The larger the context window, the better with models. Having a few TB of RAM would be exceptionally helpful.
All this just made me realise something however. Having your robot dormant and charging, is a bit of a waste. You could have robots dormant, but its compute in use to act as a compute node. If the distribution of robots is similar world-wide, we'd need a fraction of the datacenters we have now.
Using such nodes for training purposes would be beyond advantageous. And the company which can slice up the work and having training done in batches would get the big bucks. And actually, with consumer facing products soon all laden with extra ram and gpu for local compute, that applies there too.
Imagine leasing out idle time on your desktop or even laptop for cash. There may be a market here, especially with the cost of new datacenters. Any company able to securely package compute without risking data safety is going to make a mint.
> The same logic for why self-driving cars can't be cloud based, applies for robots. Something cannot be in the middle of a delicate operation and then "oops!", no network, it just stops.
I don't think you understood my post. The equivalent of self-driving is the movement control I was talking about.
Self-driving cars don't have high level logic, except for route planning. Which often is offloaded to the cloud. An extra 30 milliseconds on understanding your speech is nothing.
> Imagine leasing out idle time on your desktop or even laptop for cash.
You're missing the point. The issue is "how is the thing used" not trying to make identical break points in differing tech. A self driving car cannot have object detection, collision avoidance, human detection offloaded to the cloud. Ever. At all. A network drop can't mean it smashes into things. Or stops unsafely.
The same is true with an android. Imagine it turning on an frying pan, cooking dinner, and then going offline part way through. Or turning on a tap to wash something, and going offline while the sink overflows and destroys the house.
There are myriad of such scenarios, but local compute is absolutely, 100% necessary. Anyone betting the farm on putting network controlled devices into homes for any serious task is going to lose their shirt. Local compute is an absolute requirement, and a few TB of RAM and local compute will be nothing over the scope of this discussion (a few years minimum, just to build and kick off all these new fabs).
By the time these fabs are online, expect most smart phones to have 1TB of RAM and significant llm capable compute (gpu or other custom silicon). I would be astonished if flagship model phones in 2030 weren't sold with 1TB RAM. Note I'm saying flagship, there will be of course economy models as always. Certainly laptops will be sold in multi-TB RAM configs.
I'm not missing the point. I'm saying that some amount of local compute is necessary but the really big stuff can be remote.
You don't need terabytes to turn things back off.
Not that I want to trust "not setting the house on fire" and "not flooding the house" to this kind of model in the first place...
> I would be astonished if flagship model phones in 2030 weren't sold with 1TB RAM.
I'll be astonished if they have 50GB.
Have you looked at RAM size/price trends? I'm not even talking about the last year, just the pattern before that. We're not in the 80s and 90s anymore. The most recent price lows were roughly $3.50/GB in 2013, $2.50/GB in 2016, and $1.50/GB in 2023. If we're lucky the cheapest stuff will hit $1/GB in a few years, and the kind that would actually fit on a phone motherboard would be significantly more than that.
Samsung hit 16GB on their top model in 2020 and it's either been 16GB or 12GB ever since. Apple only went up to 12GB in the last year. Google offers 16GB. A couple niche offerings have 24GB. Why would these RAM numbers even double during the next four years?
Well there's only one reason, so you do indeed know why.
Here's the thing, by 2028, maybe 2029, 1/2 the cloud AI providers are going to go bankrupt, or severely downsize. This will likely be due to massive reduced power and processor requirements for AI. GPUs are a stopgap, and they won't be used much longer. When that happens, and the number of servers drops by 90%, along with similar power requirements, a lot of people are going to go tits up due to massive, empty datacentres and compute.
That will absolutely flood the market with cheap obsolete GPU + RAM, used servers, you name it. Current RAM prices will plummet, and those able to withstand the storm will have to compete again, not just ride the wave like they've been doing now.
RAM has been ridiculously high for a long time, it won't be by 2030.
Couple that with local SoC with dedicated compute for LLMs, and you have the perfect storm for local, on phone compute. Why this? Well, because LLMs will be in everything! Certainly in robots, autonomous weapons, and anything the crazies can shove it into. Phones won't be the unique case, it'll be sharing a hardware market with robots, and 100 other things using the same SoCs.
Local compute will be easy... RAM makes a massive difference, huge as you know, difference in local compute. And with the datacentre market in tatters, and people becoming more and more concerned about privacy, people will likely not want to whisper to a cloud LLM, about their fetishes for donkeys or moose or whatever turns their crank.
Not to mention, other aspects of their lives.
In as RAM will be super cheap, that's where I see this going.
I have a sneaking suspicion you don't agree with this assessment. I guess we'll come back here 2030 and see what's what.
What do you think the production cost of RAM is? Even if we perfectly fixed competition and ended the AI bubble right now, I don't see how new RAM prices would drop below half the previous low point by 2030. Those fabs are legitimately ultra expensive, that's why almost all the RAM makers went out of business and we're left with the current cartel.
Also "massively reduced requirements for AI, so much so that we have 90% fewer servers despite the Jevons paradox" is not compatible with "high quality models make effective use of terabytes of memory" and "LLMs will be in everything".
And even if you could perfectly reuse all the memory out of the collapsing datacenters, that's not enough to make many of the phones and laptops you're describing. We'd need a 10x increase in manufacturing on top of the massive price drops, and even if we started building those fabs today they wouldn't be done in time (and we wouldn't have enough equipment to fill them).
So... like vertical markets? Ram for the data centers, data centers for the robots, robots for the Humanoid? or... the other way around? Robots to build data centers, and robots to build Ram.
I see... tech giants and R & D at the top of the pyramid, and I have seen Gumee, the fab that started all this Korean ram prodution.
Humanoid robots that can do manual labor are going to be make or break for wealthy economies in the next two decades. Aging populations need help, and most successful nations do not have enough young people to do half the work they need done.
This is the path that Japan tried to go down and it hasn't worked out yet, but we have also solved a lot more of the technical issues since they began. going to be interesting to see if we pull it off this time.
Humanoid robots barely progressed between 2000-2020. There have obviously been incremental improvements in things like dexterity, vision, self-balancing, and locomotion, but in terms of having a useful humanoid robot, Honda's ASIMO released in the year 2000 is not crazily behind what we had in 2020. So it's not surprise we haven't seen economic dividends yet in the real world.
I think AI is what could make humanoids turn from parlor tricks to huge amounts of utility, but we're really going to have to see how it plays out in the next 5-10 years.
I think they mostly tried to go down this path before we had the transformer. With VLA models, or really now "Large Behavior Models", what's possible has changed dramatically. I've seen robot arms fold laundry now. Textile work is insanely hard, now it's just putting a lot of learned behavior together.
The current humanoid hype don't have much substances or key technologies in it, and incumbent industrial robotics companies like FANUC are already in the process of rolling the techniques created for humanoids into their robots. I personally think this is going to be just series of incremental gains for big welding bots, and nursery equipment becoming mildly robotic, like Aperture Science wall panels, than humanoids walking into retirement homes and doing dishes in the future.
Those were never techniques created for humanoid robots. Google researched using transformers as vision language action models to drive robots back in 2023 on a mobile manipulator and probably did non VLA work even earlier.
This is something people refuse to understand. The shape of the robot changes absolutely nothing about robot intelligence unless it abandons the basic concept of joints and links. Continuum robots are very difficult to control but they are also incredibly niche.
I really can't shrug off my suspicion since one of Chinese humanoid startups had added few female type body bumps on a robot and the online humanoid hype cluster("but does it do laundry or dishes" types) responded in panic with shared sentiments of defeat, that, those humanoid hype types might be just yearning for HBO Westworld bots, to state it with modesty, not even Data or Marvin who we can share drinks with. The absolute humanoid obsession doesn't make sense to me otherwise. The cookie cutter "humanoids are best poised for leveraging pre-existing human infra" response make close to zero sense otherwise.
I agree they may be strategically necessary. I’m just less convinced that “commercial humanoids by 2028” means anything close to replacing a meaningful amount of labor by then
They're pretty good at helping themselves. Close to where I used to live in Bavaria we had a pilot project of communal living for the elderly in a community of about a hundred people that included people with quite severe conditions such as dementia.
Medical and care personell routinely checked in but they were largely self sufficient and did a remarkably job of taking care of themselves, maybe most importantly the were happy and quite dignified, something I cannot imagine is the case when your only contact is a humanoid robot.
Of course in an age where every solution is yet another technology rethinking social life isn't very high up the list.
The "help" goes beyond taking care of themselves. What about food, clothing, infrastructure maintenance, and so on? An inverted population pyramid requires massive increase in the productivity of the economically active part of the population.
The elderly already have a significant regression in standard of living, often catastrophically. That's a pretty big problem I would like to improve. There are so many people who don't leave their house because they struggle to get clean and they're embarrassed, and then their social connections collapse. It's going to be a long time before we have the technology to really help with elder care, but boy is it worth it.
It's also something we could help by getting rid of zoning in cities, of course, because we constantly push low income elderly out to the edges of urban areas where they can afford to live, rather than allowing supply to keep up with demand in dense, walkable, accessible places where people who can't drive can actually have quality of life.
I agree, we could do so much better. Zoning can be a tremendous quality of life issue, and needing a car to participate in society is an expense other people in other places aren't burdened with.
Successful nations haven’t had enough young people to care for the elderly because they’ve all been employed as knowledge workers. Knowledge work is going away rapidly. The only work that will be left are things that require actual human bodies (for now).
In short, jobs like elder care are suddenly going to have an enormous number of new job seekers, since it requires human presence and isn’t taxing physical labor.
Food has gotten progressively cheaper. At the same time, both the compensation demand for farm labor have plummeted.
Demand for the output of knowledge work can increase while demand for those who produce it falls. “Knowledge work” is not a product with inherent demand, just like you and I don’t go to the store to buy farm labor.
I'm not sure I agree with the premise of that analogy. Demand for farm labor has increased, but more of it is done by machines. The number of people employed in agriculture has increased if you include agriculture technology, over the same period. Not that I'm not saying the proportion of all humans working in agriculture has increased, I'm saying the raw number has, which is what matters to people along their lifetime.
Also, both the pay rates for agricultural labor and the pay for developing agricultural machinery are much higher than the pay for agricultural labor was in the past.
Knowledge work is an ever expanding umbrella, just as agricultural work is.
The title is kinda misleading. The actual wording is not specific about humanoid robots but "physical AI" which encompasses every machinery that can be potentially integrated with AI, especially focused on mass manufacturing for the Korean case. Basically this project is about all physical infrastructures to automate high tech manufacturing industry.
I think there's a reason they're being lumped together. Memory is the reliable industrial bet. Humanoids are the speculative "what will all this AI hardware actually enable?" bet
> The wide availability of commodities typically leads to smaller profit margins and diminishes the importance of factors (such as brand name) other than price.
I'm not aware of many commodities which have only 3 world-wide sellers.
I’d say it’s more like “on groceries and a fancy dinner”. Humanoid robots sure do need RAM, both in data centres for training and in the robots themselves. :)
I wonder how Germany missed the semi manufacturing train? They had literally everything: universities, manufacturing culture, expertise and supporting supply chains, cash.
They had a large memory manufacturer, Infineon, who spun out their memory division as Qimonda which then went bankrupt [1]. They were the 2nd largest in the world at one time apparently. Looking back, it's easy to say the German govt should have thrown them a billion or two to keep them afloat. However, state intervention was very unpopular at the time in economic circles, and there was much furor over bailouts following the 2008 crisis.
Japan has an even sadder story. They were the DRAM top dog for a very long time. South Korea entirely ate their lunch.
Germany’s austerity policy after 2008 may be one of the largest economic blunders in history. It would be one thing if they merely committed self-harm, but they also used their pull in the EU to drag the rest of the continent down with them.
Reminded of Matt Yglesias’s excellent headline from 2010: Angela Merkel Lucky the Bar for “Worst German Leader” is Very High
One thing that I will never understand - how can german population not see through all that socialist bullshit she produced, promises undelivered, how much long term harm she generated across whole Europe (not just EU). Hard push for (now visibly) failed immigration at all costs, nuclear energy rollback (and she was nuclear physicist at least by her studies) and subsequent buying of foreign coal-based electricity, fragmentation of EU.
She literally licked putin's boots well into Ukraine war and still thinks licking his ass is the correct course to solve war in Ukraine (which started 2014 and it was pretty nasty already back then, the world just didn't care also thanks to her).
She is by far the biggest catastrophe modern Europe encountered after Hitler. She helped remove any proper fighting chance for the top dog Europe had for 21st century. She singlehandedly caused proper hate against EU in large parts of (not only) eastern EU population, and hence the rise of populist left or right wing politicians whose whole success story was just point at her failings and criticize, enough to get 20-30% of the votes and even win elections, repeatedly. She literally made people like Orban or Fico.
She still admits no mistakes, even wants to become german president. How effin' out of touch with reality she is.
... what? How's that the first thing that comes to mind about her, before "neoliberal", "conservative", or "austerity"? For that matter, when has the CDU ever been anywhere near socialist, in Germany or in the EU parliament?
100% agree we're still dealing with the fallout from her policies though.
Yes people voted for her, she is still very popular. Doesn't change anything I said about her negative impact on entire continent that is extremely visible.
Maybe she did tons of good so it somehow averages out, but I certainly haven't heard about it (now is your time to defend her), everybody saw consequences of her disastrous policies that affected entire bloc.
This is untrue, entire eastern EU screamed like crazy against such immigration when she started importing immigrants en masse, "wir schaffen das". France faced the same with entire communities from Maghreb, most didn't work, 0 integration even after a generation or two spent in Europe, they knew language only if it was also their native one. Nobody in the east wanted that, entire political blocs before elections formed on opposing this and other fun EU moves (remember banning regular lightbulbs without good-enough LED replacements in entire EU market? Pepperidge Farm remembers... Or cutting down most of Borneo ancient rainforest to have effin' EU-subsidized bio fuels that ruin car engines quicker? Same story, I know this is EU but Germany is by far the strongest voice in and these are typical merkel moves - big moves regardless of consequences and picking up pieces later).
East had no prior experience with migrants due to living for decades in effectively prison camp guarded by soviets, no travel or other exposition to other ways of life. We were very monolithic cultures (and still mostly are).
The voices were completely ignored and overruled by behemoths like Britain, France and most powerful voice in the bloc - Germany.
Don't revision past, I lived through it, saw masses of people getting absolutely mad pissed off and feeling helpless and unheard with arguments which over time proved them mostly correct.
I don't hold those opinions myself, some of our best friends are muslim immigrants but oh boy go to the eastern EU without camera and ask random folks on the street outside capitals, or just listen to them. Or look at election results, this gave russians and their constant influence very good arguments since they position themselves as 'guardians of traditional family values in society', regardless of how its true or not (clue - its not but thats detail here).
> Looking back, it's easy to say the German govt should have thrown them a billion or two to keep them afloat. However, state intervention was very unpopular at the time in economic circles, and there was much furor over bailouts following the 2008 crisis.
Absolutely, Germany essentially abandoned its position in industrial leadership solely due to neoliberal ideology. Just compare the trajectories of Germany and China in the last 20 years. One country planned and implemented a proper industrial policy, the other hummed and hawed about the infallibility of the market and thus essentially just gave up.
There were plenty of other DRAM manufacture before them as well. But the Boom and bust cycle wipe out those who weren't big enough or adequately prepared.
Nearly every single DRAM cycle we go through the same thing. People are saying memory are so cheap we are going to get 32GB/ 64GB / 128GB very soon. Just like they did 2-3 years before. Then every single DRAM up cycle they were earning insane amount of money, with this one being exceptionally high. Not only covering their previous lost but also extra profits. And every single time people cry out price fixing.
And every single time, Interviewer, financial reporters or share holders will ask TSMC's CEO Morris Chang the same question, and this is before TSMC was even well known in Tech cycle, will they make DRAM given it is just Fabs, TSMC could do it. And every single time in the past 20 - 30 years Morris had to explain the whole thing again.
Frankly speaking I have been rumpling about DRAM being commodity on HN for nearly 20 years. It is the first time I have seen a few comments finally picking this up as shown in the current top comment.
I don’t remember any time in recent history where demand skyrocketed so fast and so far away from supply. In previous cycles you mostly had supply driven shortages (remember all the fires?).
Maybe crypto surge was the closest to what we observe today? But again completely different magnitudes of demand. No company put 1T of capex for mining
Realistically, when it comes to the semiconductor market, there aren't many viable options outside of East Asia. I don't mean this in the sense that East Asians were somehow "chosen," but rather that the semiconductor industry inherently requires a large number of highly educated employees working together. The problem is that the working hours inevitably end up being very long. If you actually go work at one of those facilities, you have to wear a "cleanroom suit" (bunny suit), and it's physically demanding. What I'm saying is, you need highly educated personnel who can be mobilized at any time when a problem breaks out in the middle of the night, and who can be hired at relatively low cost. East Asia has a massive educational infrastructure — schools are very large-scale and the system is extremely well-developed — making it hard for other regions to compete. And indeed, the average working hours in countries that do semiconductor manufacturing are extremely long
In other words, it's an industry where you have to grind white-collar workers as if they were blue-collar laborers.
I’ve always wondered what is unique about semiconductors that PhDs need to work like assembly line workers. I’m sure they’re not solving partial differential equations all day, but what’s so different between different batches of chips?
The industry inherently deals with extremely hazardous chemicals, and on top of that, during semiconductor production, there are many things that have to be recorded and tracked.
A lot of the processes are automated, but at the points where automation hasn't reached, there are quite a few things that are genuinely complex to handle.
I think it’s more like highly skilled technicians, to scale up. Plus PhDs and other scientists to do the simulations and analyze the data for new designs.
Answer: they don't, they just work them into the ground because they can.
Mainland China also has the 996 schedule for office workers purely as a cargo cult ritual, forcing people to sit at a desk at midnight and pantomime doing work.
You're not wrong, but from what I understand, the issue really comes down to on-site personnel and supervisory staffing.
As for the 996 culture, I agree to some extent. My Chinese friends hated it too. But in China, there's this thing called neijuan (involution / the rolled up scroll).
there are just so many job seekers that people are forced to endure it. What neijuan means here is "Eating one's own flesh" basically knowing that this competition is damaging to everyone involved but doing it anyway.
The people in the fabs aren't PhDs, they're extremely skilled technicians. They may have a similar number of hours of experience/training as a PhD, but that training is in troubleshooting automation systems, not doing research.
I believe some of the earliest Intel fabs were in New Mexico (Shiprock and Rio Rancho). What combination of the above did New Mexico have?
When New Mexico and Germany had fabs, South Korea was still a developing country ruled by a brutal dictatorship.
What happened was simple - both Taiwan and South Korea and now China took concerted steps in investing into their semiconductors businesses. South Korea did this indirectly through favourable arrangements for the industry players via the chaebol system, while China and Taiwan did this with more direct government investment into the industry.
Sure, you can't just dump money into the industry and become a semiconductor player, else the Middle Eastern countries would have tried that ages ago. Yes, the talent being locally present is important but you're once again bringing up tired tropes about Asian working culture as being relevant.
I'm not saying Asian culture is the main factor. Yes, it's true that authoritarian governance driven by dictatorial regimes and chaebol politics has played a strong role, but fundamentally, the long working hours are simply inherent to this business.
You brought up the New Mexico story quite well, but that place is notorious for the exploitation of Navajo women's labor. In the first place, the factory was occupied and shut down by the American Indian Movement. You know full well that this is a story about the exploitation of Native Americans, so why are you bringing it up like that?
The history of Shiprock itself is, at its core, a history of "cheap, obedient labor." You frame it only as state-led investment, but the reality is that the culture behind it is complex.
What my post is pointing out is not that "Asian culture is superior." What I'm pointing out is the harsh working conditions in Asia — where working hours are extremely long, and even highly educated workers are inevitably subjected to grueling hours. Why do you think TSMC's Arizona fab in the U.S. keeps getting delayed? The U.S. invested money through the CHIPS Act, but American engineers refuse to accept the "military-style 24/7 on-call readiness and brutal shift work" that exists in Taiwan. TSMC founder Morris Chang himself has pointed this out before.
What I'm saying is that the educational infrastructure is so well-established that it's easy to produce a large supply of highly educated workers, and that these highly educated workers then have to be submissive to inhumane working conditions. This isn't about Asian superiority — it's actually pointing out something bad about Asia. But from the context of your comment, it seems like you misunderstood me as saying "Asian work culture is superior" and replied based on that assumption. That was never my intention.
Before you leave a comment, I'd ask you to show some basic respect to others.
This is extrapolating from a single example of something that has worked and the conflating correlation with causation.
There are plenty of places with highly educated cheap workforces who work hard. Eastern European culture is almost identical down to the whole "tiger mom" stereotype.
The US is full of the "military-style 24/7 on-call readiness and brutal shift work" - at the high end silicon valley is built on this, and at the low end every single non-unionized factory is this.
TSMC has never built a fab outside Taiwan. Of course there will be problems.
There were probably many complex factors at play. Personally, I think the biggest one, as TSMC's Morris Chang said, is the inhumane working conditions imposed on highly educated workers. But there were likely also issues around permitting and regulatory procedures, as well as the overall cost structure.
As you said, if it were just about labor, other countries would probably have some supply of it as well. But in the case of Eastern Europe, there was likely American pushback against the European continent. As you know, semiconductors today can't be made entirely by a single entity. They're connected through a chain of trust. If Europe were to move beyond just producing semiconductor equipment and start directly manufacturing semiconductors through fabs, it would easily become a competitor to the U.S. rather than a supply chain partner.
In fact, the semiconductor chain is deliberately fragmented so that no single player can monopolize it.
On top of that, the U.S. is using South Korea and Taiwan to contain China. Under the ideology of protecting foundries from Chinese aggression and industrial attacks, the U.S. is sending the signal that it can cut off the supply chain. Eastern Europe, on the other hand, is tied up with the EU, making it much harder for the U.S. to control.
In the end, what matters when the consumer nation, the U.S., outsources production is how securely it can relocate it. Look at what happened to Japan's semiconductor industry. It was crushed through the 1986 agreement. The U.S. simply does not tolerate the emergence of an independent manufacturing hub that possesses sovereign economic power.
What matters is whether the U.S. can maintain control while keeping the price low.
Feels like the real clash happening here is that the reality is suggesting that the values of mean educational level of the bottom 99% of workforce outweighs that of the top 1%, and that being uncomfortable to some so much so that there has to be something else. But isn't that just it?
There's a story in one of Feynman's memoir where he figures out that pausing the live system and debugging its physical RAM stack is turning out to be more time consuming than simply scheduling a new corrected task, on some particular 1940s mechanical supercomputer he was assigned to as a tech. It might not have taken Feynman to notice that, but you can assign Feynman for that, and it worked for the Manhattan project.
The parent comment isn't (just) reiterating the tired tropes, but pointing out that East Asia has an "educational base" similar to industrial base that supports its high tech. I don't think that much is so strange way of thinking. The state of ME countries(maybe except Iran) soft proves it - they don't believe in such a thing. And they don't have a semiconductor industry. Pure coincidence? I doubt it.
(And on "This isn't about Asian superiority — it's actually pointing out something bad about Asia." from jdw64, yuuup 100% it is quad plus bad - IMO a thing about East Asia is that there's zero inter-national mobility due to the notoriously high language barrier, so competitions are closed to within borders, and the bar just drift skywards indefinitely because of that. There was a massive domestic hiring freeze in Japan during the 90s that made "janitors with a PhD" actually not so rare, but none of them hit the global labor market or started companies - the Japanese bar for janitors just went up to PhDs. It is said that success of Japanese 7-11 was partially attributable to that event, that, when you happen to have all the cashiers manned 24/7 with top scientists, you can just throw million different tasks and they can handle it perfectly, put aside whether they're happily doing it)
New Mexico has (pockets of) exceptionally well-educated population. 3 of the top 100 counties for PhDs per capita are in New Mexico (including #1, Los Alamos).
That's precisely my point. You need both the talent and the investment to execute this. Simply stating "Asian culture" is a stretch at best. The Asian economies made a ton of investment into semiconductors from the 1990s onwards, which are paying off now. Los Alamos is a pocket of exceptional PhDs because of the US government's investment into the institutional projects as well as building the infrastructure to attract talent to come and work in the middle of nowhere. That ended up creating a broader ecosystem of innovation there. What has the German government done, except for creating policies and regulations to ensure incumbents remain powerful? India pumps out more STEM graduates than most of the world combined, but what does it have to show for that in semiconductor technology but for a few middling projects?
Los Alamos would still be a desert without USG giving it a few billion dollars every year.
Chip fab locations have traditionally had more political than economic importance. Matrix multiplication chips and RAM have been the recent exception, while TSMC has long been the geopolitical exception. ASML's location only matters to the extent that it gets ordered not to sell to someone.
According to late Intel founder Andy Grove, having domestic manufacturing, including chip fabrication, is very important for a country's ability to innovate.
Not just ASML, but Zeiss the optics company is German and optics are one of the most important parts of lithography.
Regardless there are fabs in Germany, I know of at least 7, most located in Dresden. They mostly focus on older larger nodes, since for the German auto/industrial sectors that is all that is required. It's a much bigger industry than other European nations, especially like the UK which only really has one main fab in Newport.
IIRC Taiwan took a page out of Singapore's playbook and went all in on electrical engineering and adjecent fields. It was very much a long-term strategy. Germany probably didn't feel nearly as much pressure, and was already very strong in all industry.
Intel was supposed to build a fab in Magdeburg, which would have been great, but apparently the reason it was canned (2025) was they couldn't secure enough customers.
Memory has only really recently become lucrative. Germany still has heavy machinery, trains, drilling machines etc all of which will be needed for a long time regardless of whether the "bubble pops" or not.
Heavy machinery is starting to. Computer vision for robots is a big deal, and takes quite a bit of processing power. Robotic mining, earthmoving, and even construction equipment is exactly where Germany will innovate. Not to mention drones - Rheinmetall needs DRAM...
same thing that happened to Japan/US: they financialized their economy, companies moved from making things to pushing papers and outsourcing the making to China
> I wonder how Germany missed the semi manufacturing train?
Socialism or more its german variant, a system that can spend much more than more capitalistic power holders can ever earn and doesn't really plan well for future. Just look at it - very protected jobs, stifling bureaucracy, very hard to fire people, brilliant folks are definitely not compensated accordingly compared to (below) average peers - more often than not they earn the same. Its not agile economy nor workforce by any means, in contrary.
The feeling that the German Way (TM) is The Right Way, regardless of situation. If it worked in the past, it will in future, right. That leads to stagnation, complacency and when competition leap frogs them by the mile, surprised puzzled looks and wondering how it all happened.
This is how overcapacity happens in a commodity. The is a rush to expand capacity to meet demand, and there will be overshoot. I hope these factories are built quickly so the memory crunch eases.
if you look at the price per gb of memory, at today's prices its only equal to 2014. People don't complain about memory prices back then. The reality is modern world just need much more memory. Before 8GB memory was enough, today my 32GB memory is 95% used all the time due to chrome. And my 2TB SSD is filled with 4K HDR videos. This is not counting massive storage needs by datasets, AI models, KV caches, and bandwidth needed for LLM inferencing.
Why humanoid? Surely there must be a superior physical form factor than one mimicking human anatomy. Is it just supposed to be more psychologically acceptable?
> Why humanoid? Surely there must be a superior physical form factor than one mimicking human anatomy.
There probably (certainly) is. But if you want to build a multi-purpose platform, you’ll soon be faced with a dumb challenge: nearly all interfaces (door knobs, taps, electric switches, cutlery, sponges, every single button out there, pillow cases, wrenches, hammers, signs…) are made for humans. Placed at human hand level. At human eye level.
Nearly all environments (houses, streets, sidewalks, factory floors, offices, toilets, bathtubs,…) are made for humans. Wide enough and tall enough (or short enough, for bathtubs) to accommodate human bodies.
So until we can find one or more form-factors superior enough to justify we adapt everything around it or them, betting that the easiest way to build a single multi-purpose platform able to do most things (and not n platforms for n+ use cases) is to borrow the shape most things are made for wouldn’t surprise me. Plus, you get a wider market.
And then, once you have happy-ish customers, figure out which of these human attributes and shapes aren’t actually needed to do the job.
Never said it was interesting. Just that it’s easier to sell tech that fits your customers' environments than to get customers to overhaul their environment to fit the unproven tech you want to sell them.
Most fixed infrastructure like switches and buttons that can be operated with a single hand are amenable to any robot with at least one hand. We've had Sawyer demonstrations operating tools clearly designed for humans for more then a decade ago and Mobile Aloha had demonstrations showing how it can operate the switches of an elevator just fine. None of these are humanoid robots.
The moment you have mobile tools, whats the point in forcing the robot to hold them using a human hand? You can put them on a tool changer now or have a gripper that works for the specific task. Why does a robot need to hold a wrench using a humanoid hand?
>Nearly all environments (houses, streets, sidewalks, factory floors, offices, toilets, bathtubs,…) are made for humans. Wide enough and tall enough (or short enough, for bathtubs) to accommodate human bodies.
Uhm, now we're getting into stupid territory. All of those environments have flat floors. Flat floors are not an environment that are exclusively built to accommodate human bodies... The flat floor is designed for ultimate flexibility. It can be used for anything. Furniture, wheelchairs, wheeled robots, furniture on rolls, animals, and also humans.
All of the environments you've listed should preferrably be wheelchair accessible for disabled people (in terms of locomotion at least).
>So until we can find one or more form-factors superior enough to justify we adapt everything around it
Is this some kind of joke? Factories already make heavy use of UGVs and stationary robot arms and build custom end effectors for them. It's also an extreme strawman to suggest that wireless/electronic interfaces require finding a "superior form-factor" to the point that it feels insulting. There's also often an easy wheelchair accessible equivalent. E.g. a button to activate the electronic door opener at wheelchair level can still be at a comfortable height for standing people.
>And then, once you have happy-ish customers, figure out which of these human attributes and shapes aren’t actually needed to do the job.
So solve the impossible (come on you know it's hyperbole) first, only then can you build a simpler system.
I think its incredibly unreasonable to suggest that you need to solve every single problem ever encountered in human existence to be allowed to solve one much simpler problem.
I’m not arguing that humanoids are the only robots that can operate in human environments, nor that specialized robots shouldn’t exist. We already know they should, and do, and work very well.
I’m saying it depends on what you’re after.
If you want a specialized machine, then by all means optimize the form factor for that task.
If you want a general-purpose platform, then you’re faced with a near-infinite variety of environments, tools, and situations: homes, offices, hospitals, factories, hotels, streets…
Almost all of them were designed around human reach, dexterity, height, mobility and perception.
> The moment you have mobile tools, whats the point in forcing the robot to hold them using a human hand? You can put them on a tool changer now or have a gripper that works for the specific task. Why does a robot need to hold a wrench using a humanoid hand?
It doesn’t. The wrench was just one example among many. Nearly all our tools and interfaces are designed around human hands and bodies. If a different manipulator works just as well across all of them, great. My point isn’t that it must literally be a human hand.
> I think its incredibly unreasonable to suggest that you need to solve every single problem ever encountered in human existence to be allowed to solve one much simpler problem.
That’s not what I’m suggesting.
I’m saying it’s a deployment strategy. Start with the form factor that’s already compatible with the largest installed base of tools and infrastructure. Not because it’s mechanically optimal, but because it minimises the amount of adaptation required from the market.
And enables you to sell to markets you aren’t even aware exist.
If, over time, you discover that hands don’t need five fingers, legs don’t need knees, or an entirely different morphology delivers much better performance, then you have both the experience and the demonstrated value to justify changing the robot, or even convince your customers to change the environment to fit a better robot.
They’re basically trying to go for the "one size fits most" of robotics. And yes, we both know how well that usually fits anyone.
I’m not convinced either, but I can understand the logic.
There are just a few reasons - humanoid make sense, mostly for multi purpose tasks - where if you want a robot to be multi-job, do almost everything a human can do at work --
If you want a weld you need a 1 arm robot, if a robot to weld, then stack, then push parts on a cart across the factory - then sweep up, then etc.. etc.. perhaps a humanoid is alright.
There will definitely be too many people comfortable with ownership / master relationship with a humanoid robot that will do their bidding.
For multi-purpose tasks, human hands and arms are excellent, but only 2 are too few for many tasks. Humans very frequently need to have specialized gripping tools in order to accomplish tasks that cannot be done with only 2 arms, or they must work in pairs.
A good mobile multi-purpose robot should have 3 arms, or 4 arms for symmetry.
Human legs are normally not necessary. A mobile robot would just need some means to raise and lower its wheels, so that it could step when ascending or descending stairs.
A human head is not useful. The place for the "brain" of a robot is in its "chest", because robots do not have the limitations of living beings, where the very slow propagation speed of the nervous signals forces the nervous systems to be concentrated in the proximity of the main sense organs.
Instead of a head, one should have a couple of mobile arms with video cameras at their end, somewhat like the mobile stalks of crab eyes.
Of the components of a human, only the hands and arms are models useful to imitate. Cephalopod-like arms would be even more versatile than human-like arms, but it is likely that they would be much more expensive at similar performances.
Having the size of a human and human-like hands and arms is good for working in environments designed for humans, but having the shape of a human has no purpose.
I understand the argument but its honestly ridiculous in my eyes. How about a set of arms that can reach into dishwasher and stack dishes and a washer/dryer to fold laundry... Except even without solving the bipedal movement, that doesn't exist at a consumer price point.
Why are we pretending the hardest version of this is close to existing?
It doesn't need to be at a consumer price point first, it needs to replace a human at an existing warehouse or manufacturing role first, and that's achievable in the next two years at this point.
When you have arms that can reach into the dishwasher, you're also going to want them to put away your dishes. And so suddenly they need to get up high. And you're not going to have a SECOND set of arms at your washer/dryer to fold laundry, you're just going to buy a second DLC for your existing robot. And it needs to get between those places, so if you have stairs, wheels don't cut it. You need a bipedal robot very quickly.
I am familiar, I'm a big fan of Dean Kamen's work. So far, we haven't seen a single wheeled stair climbing vacuum cleaner, even though the original iBOT is 23 years old.
That solves the horizontal mobility problem. And then you have cabinets - and wheels don't solve the vertical mobility problem. So then you need a scissor lift on those wheels, or a hydraulic lift.
The robotics nerds always end up back at bipedal because it's vastly simpler once you're already solving arms.
A scissor lift or some other thing seems a lot simpler than trying to allow for a bipedal robot to knee to reach things close to the ground. Or if legs are the best option, why not a quadruped? Much more stability with a lot of the same advantages.
Use wheels and buy two if you have to...the Roomba solution. Besides, why do you need to solve stairs the hardest way possible, a fully bipedal robot, before it moves past vapor?
I think the key is that none of our actual home use cases can be done with just arms. You don't need your folded clothes sitting in front of your washer and dryer, and a set of arms can't handle folding sheets.
Why not? I would love to have a set of arms that could flip the laundry from the washer to the dryer and then take it out of the dryer and fold it and put it in the basket.
I understand that particular use case sounds cool! I really do.
And then you want them to put away your dishes, and they can't, even though it's just a software update, because they're across the house. And they're BIG, so you don't have room to store two anyway.
Back around the turn of the (20th) century, electric motors were expensive. It was not uncommon to buy one motor that could do multiple things, like this vacuum/grinder/buffer/blower/pulley:
If we start making robot arms at scale, they're going to get cheap.
I'm also not sure people are really going to want bipedal robots walking around their home, blocking the hallways, recording you in your underwear, etc.
Sure, an electric motor is like 10cm on a side. A set of robot arms that can fold laundry are like a 100cm cube. Most people aren't going to have space for two of them.
A washer+dryer is pretty huge already. Seems like it could get some robot arms without changing the form factor dramatically.
That said, I think this is way farther off than anyone thinks. I want to know what the maintenance schedule looks like for robot arms. Looks like a lot of small moving parts. Probably a lot of plastic gears.
In an industrial setting, sure, maintenance is just an expense. But wheels require less maintenance and factories can be designed around the robots.
As I said, maybe earlier in this thread, I've now seen a laundry folding robot work, but I think loading a dishwasher and then putting away dishes is going to come first.
Ok, so your idea is to sell a machine without software and then hand wave the software part away by saying it will be released at a later date? Sounds like a scam to me. You'll just end up with another AMD situation where the company is only interested in selling the hardware and has zero interest in developing the software, because it costs money and can only ever generate revenue by selling more hardware, which will make the hardware focused company feel vindicated in deciding to not put effort into the software.
My idea? No. I'm pointing out that if you release something capable of loading a dishwasher and putting away dishes, it could be as little as a future software improvement to make it do something else.
Human spaces are built for humans. Outdoors cars and quad coppers are a great form but constrained by stars, doors, and low ceiling makes them a poor fit.
Alternatively a 2 foot tall or a 20 foot tall humanoid robots aren’t particularly useful. But a good enough 5-6 foot tall humanoid robot can be swapped into an assembly line wherever a human is currently working without redesigning that workspace.
>The most costly of the megaprojects involves a commitment by Samsung and SK Hynix of $585 billion to build new chip fabrication plants..... The government’s goal is to double South Korea’s production of dynamic random-access memory (DRAM) within five years.
"Double"
I was wondering about that $585B number only to read they plan to Double the capacity. The two firm together is nearly 70% of the current DRAM market supply. I guess we are really going into an era where future Phone and PC will have 16GB as minimum, possibly more.
Even if the market are back to normal, that is doubling the revenue. The current AI scale and Cloud infrastructure buildout is simply beyond what I could comprehend.
It's from the president's speech. Too lazy to look up the actual text but I guess he meant "pillars", a common metaphor in East Asia. In English axis and pillar are distinct but in East Asia the line is blurry.
For example, the Japanese word 軸 (jiku) is used to mean the "axis" of a graph, but it is also used in business to mean the "core pillar/backbone" of a strategy (e.g., 経営の軸 keiei no jiku, literally "the axis of management," but conceptually "the pillar of management").
The speech was delivered in Korean so this is a choice by a translator. I don’t speak Korean but I asked an LLM and it says …
the phrase used is "대도약" (daedoyak), which literally means "great leap forward" or "great jump forward." This is NOT "대약진" (daeyakjin), which would be the direct translation of China's "Great Leap Forward" (大跃进).
To expand a bit, even saying 대약진 _daeyakjin_ "great leap forward" wouldn't have turned many eyes, because _dae_ is just a common prefix ("great") and _yakjin_ is also a common word meaning "leap forward, push forward, improve". The word simply doesn't have the same connotation of the English phrase "Great Leap Forward", which is almost always used for the infamous Chinese movement.
If a Korean speaker wanted to talk about that Chinese movement, they'd use the full name, 대약진운동 (大跃进运动): the great leap forward movement.
Better spend it now, people won’t need greater than 1.5tr parameter models
and battery powered consumer devices will be able to run those and lower sufficiently capable models by then, distributing the need for compute away from capital projects
the glut will be enormous
yes, immortalize this phrase just like the 640kb ram phrase, I’ll stand by it
Sure, but I'm not sure humanoid robots is the best form factor for this. E.g. something with a wheeled base makes movement calculations a lot easier since you don't need to deal with balance while moving.
(Fwiw, >20 years ago RethinkX correctly projected the exponential cost declines of solar and batteries, when everybody else was drawing straight lines.)
> Humanoid robots will enter the market at a cost-capability of under $10/hour for their labor, on a trajectory to under $1/hour before 2035 and under $0.10/hour before 2045.
I dont see it, unless this is an expectation that a robot will work for 50 years without maintenance at capex.
Why doesnt a comparable tool, like an excavator, work with this math? Why arent they 100 times cheaper to run than 20 years ago? Excavators can cost 50 - 100k pa in maintenance and fuel costs.
Why does creating a multifunction tool, with even finer tolerances, working in human safe workspaces cost less?
Because supposedly they would repair themselves or be repaired by other robots, and energy would cost less and less, anything would cost less and less if work is increasingly done by robots than can be improved year after year.
It reminds me of predictions for the cloud, that someone would just write a perfect cloud formation script and then be able to sack all of their employees. But instead it just updated a lot of System Engineers to Cloud Engineers and Devops positions.
The direct implication is that not only will robots definitely be allowed into human spaces, and definitely be able to use generalised hardware to perform fine detail technical work but also contain only servicable/replacable parts.
I dont see any of that as proven. I think its far more likely that a very limited subset of robots will be allowed to be near humans at all, and insurance for them is going to be extremely expensive. I think if I had a look at Australian law they would be considered Plant, guaranteed if they have any pressurized components or strong lift capability. Maintenance, is going to cost an arm and a leg, and the units themselves will be largely proprietary, so you are looking at ongoing maintenance via the vendor, much like what the tractor industry is working on.
But this logic applies to all machinery and technological progress. In the end it boils down to cheap energy and materials. The shape of the machine is actually irrelevant here.
Your "Fwiw" feels meaningless and hollow. The fact that there are stupid people out there, doing stupid things does not make you a savant just because you didn't make the same mistake.
This person clearly hasn't spent much time educating themselves how high volume manufacturing works, nor have they spent much time on flexible manufacturing systems either.
The entire article boils down to "Humanoids! Humanoids! HUMANOIDS!"
All current jobs have human input and output interfaces. If you want to sell new technology then it will be easiest to accommodate the already existing infrastructure.
They can sell “employees” who don’t require salaries, onboarding, healthcare, 401k, benefits, etc and then leave after two years of being lazy and try to sue you. (This is how it will be marketed)
Economies like South Korea and Japan have a drastic population deficit that means there are simply not enough people around to perform many kinds of manual labor tasks.
Sex bots and disposable police. This is basically the future in every dystopian SciFi these politicians and oligarchs grew up watching. This is just living out fantasy.
Depends. Still bottlenecked by ASML I would assume. Otherwise the current record holder TSMC manage to bring up a Fab operation and HVM in less than 18 months. Of course that requires a lot of planning before hand.
>We make non-humanoid robots because they're better at doing things than humanoid robots are.
They're only better at some things, and some they can't do at all. If there's going to be a mass market for household robots, people are going to want a robot that can clean their toilet, read a book to their kids, unblock their kitchen sink, and climb a ladder onto the roof to get leaves out of their gutter, not a separate robot for each job. Any non-humanoid robot would struggle to do everything around the house that a human can do.
I generally don't like the phrase cargo culting, but building a humanoid robot is probably the ultimate form of cargo culting.
The idea is this: You build a robot in the shape of a human with the hope that by building the robot well enough in the image of humans, it will become sentient and intelligent on its own.
That is literally the pitch of every single humanoid robot company on the planet.
South Korea is facing a serious demographic crisis, in the not too distant future it'll be a country of mostly elderly folk. I'd be interested to know if this investment has anything to do with this, since robots may be needed in the absence of young able bodied folk.
It's crazy how much 1at could improve QOL for their citizens and also improve and diversify their economy. Alas, they're just going to subsidize ram prices for everyone when this current cycles goes from boom to bust.
Since this money belongs to Samsung and Hynix, it cannot be used for charitable activities. However, it is much better to build new cities, semiconductor factories, and power plants than to pay dividends to shareholders. The construction industry is one of the easiest ways to stimulate the economy.
What's incredible is how much resistance there is in the U.S. to do what for other countries is the obvious strategy forward. The U.S. — after decades of seeing manufacturing being outsourced — suddenly has an incredible advantage in data centers that is producing onshore facilities that are adding hundreds of billions of dollars in annual export revenue, and instead of there being a united front to maximize that advantage, there are huge obstacles being thrown in the way of the companies, administration and the state governments leading the data center expansion campaign, with Sanders and AOC calling for a national data center moratorium.
I live two miles (as the crow flies) from the site of a new construction of a data center. It’s being built by a European company. They will be installing gas turbines for power generation. Are we far enough away so that low frequency sound won’t affect our quality of life? We are seriously considering selling our home and moving. It’s a huge gamble for us. We are also noticing homes are staying on the market much longer than previously.
These data centers also get huge tax incentives because “jerbs”. Local politicians don’t understand after construction they don’t bring jobs or revenue to the city or county. This is where the resistance comes from.
That's a reasonable concern. Any industrial site is going to be a concern for nearby residences. But even proposed sites far away from any residential settlements are being opposed. The proposed National Data Center moratorium, for instance, is completely sweeping and would be devastating for U.S. strategic interests.
As for tax revenue, data centers contribute significantly to local communities.
"Loudoun County, Virginia, often dubbed the "Data Center Capital of the World," provides a compelling example of how data centers can reshape a community's fiscal landscape. In 2018, the county hosted about 13 million square feet of permitted data centers. By 2024, that figure skyrocketed to 43 million square feet—a 231% increase in just six years.
This remarkable growth has substantially boosted Loudoun's tax base. The county's data center industry now contributes an estimated $890 million annually in tax revenue, nearly matching its entire operating budget of $940 million.
What makes data centers particularly advantageous is their cost-to-revenue ratio. For every dollar of tax revenue received from data centers, the county spends just $0.04 to support them, compared to $0.25 for traditional businesses. This financial efficiency has allowed Loudoun County to maintain the lowest real property tax rate in Northern Virginia—approximately 25% lower than neighboring counties."
this is what a country in decline looks like. When people oppose any sort of progress. Then will come acceptation or rationalization like the europeans do where they pretend they dont turn the AC to be enviromently friendly instead of just admitting they dont have the money for it
The title sounds to me like: I am going to spend $1000 in groceries and dance lessons. That is, two very different things lumped together.
Memory chips are like groceries, essential commodity parts, a no-nonsense investment. Humanoid robots are like dance lessons, it is cool, it is sexy, and it may pay off in the future, but the value is much less certain.
$585B on new fabs, $357B on AI data centers, and $5.8B on humanoid robots. One of those numbers is not like the others
I don't have the historic numbers at hand, but I would assume that for each of those categories this is a similar proportional increase, so it's similarly notable to mention the increase in spending on humanoid robots.
How can you spend more on a still early stage tech, well in the depths of R&D?
Androids (humanoid robots) will require loads of ram, and loads of model training under the current paradigm. So it sort of makes sense. At least, I see robots as the top of the pyramid.
Autonomous (non-teleoperated) humanoid robots that can do useful work in an unfamiliar environment do not exist. And nobody's close enough to making them to understand if they're possible with our current level of technology, let alone how.
Most initial work for them would be in familiar, well-controlled environments - replacing humans in existing factories. I think whether they'd be cost effective for that will remain unknown even after a few years in service though.
Factory work that can be straightforwardly automated by robots generally already is, by special-purpose factory robots. The remaining tasks are generally:
* Tasks where poorly-paid humans are cheaper than expensive factory robots. Humanoid robots are more complex / fiddly so will be more expensive than existing factory robots. No help here.
* Tasks where human dexterity has an advantage over state-of-the-art robot actuators (e.g. sewing fabric panels into garments). Better robotics could help here, but the advancement needed is better actuators, not AI and a humanoid form-factor. And if you solve this you'd be better off putting your new end-effector on an existing 6dof platform.
* Supervising robotic equipment and handling exceptions. But then you get to "handling poorly-specified unfamiliar tasks in the physical world" which is not currently a solved problem and there's no guarantee just throwing more compute at it will be sufficient to solve this. So far all humanoid robot demos have either been either known tasks in a tightly-controlled environment, teleportation, or so poorly-functioning as to be obviously not fit for purpose.
Totally agree.
I'd also point out that what makes this whole thing smell of being a grift is the fact that what's being chased is humanoid.
Humans are not the pinnacle of dexterity or stability. And freed from biological constraints, it makes no sense why we'd use the human form as a reference.
Making a humanoid robot is a hard thing to do, for sure, but it's also not particularly useful. The routines needed to balance a robot or correct for a slip are interesting to solve, but don't really make for a better robot which is more capable of doing the dishes or folding the laundry.
If I'm amazon, for example, then the most useful form factor for a general purpose robot is 2 arms on a 4 wheel omnidirectional rolling platform.
Our hands are actually amazing, to a degree that’s hard to comprehend unless you’ve tried to build something equivalent. Our finger tips can discern features down to around 10 microns, detect a wide range of vibrations and single shocks, manipulate objects two orders of magnitude smaller than them, grasp an incredible range of weird shapes and materials reliably, repeat pre-learned motions remarkably quickly, and with training carry kilograms, all while weighing about a kilogram and reliably operating maintenance free for tens of thousands of hours. Oh and they’re made of meat.
An artificial manipulator with even half of these features would be the holy grail of robotics. Doing them all at once is a miracle.
Completely agree. My point is more that even though our hands are amazing and even if we decided we want all the features of hands, there's no reason why you'd design such a device with 5 fingers and an oddly placed opposable thumb. There's no reason you'd have more than 2 joints per finger.
That's what I'm calling a grift of the humanoid robots. They are focusing pretty heavily on making manipulators shaped like hands, which don't perform anywhere near hand capabilities. The form is mattering more than the functionality which introduces a lot of unnecessarily hard to solve functionality problems.
The actual hard part of doing something like folding cloths isn't even the manipulation (though it's part of it) you mostly just need a few pincers. The hard part is that fabric changes like crazy in a 3 dimensional way on every motion. Just identifying "this is a shirt" is a hard problem to solve. Further deciding how to fold said shirt is extra difficult. But even further, getting a robot to know "this is where this shirt should be put away" and "this is how I should handle new cloths" is crazy hard. That's the part that I've seen absolutely no evidence that any of these humanoid robots are making any sort of progress on. The most impressive ones are cheating with a guy in the room over doing the task. Not exactly something I'd want in my home. Perhaps in a nuclear reactor.
I would say that turtles are the pinnacle of stability, and longevity, and so we have Roombas. But, Racoons are at the pinnacle of dexterity, and Octopodes are at the pinnacle of dimensional manipulation...
but personally, I would welcome our cyper-coon overlords.
We’re experiencing gpt-2 moment in robotics now. This means in about 2-3 years they will do useful work (cooking, repairs, cleaning, etc).
The extrapolation cannot be justified. It may be much longer or tomorrow.
[dead]
So far all humanoid robot demos have either been either known tasks in a tightly-controlled environment, teleportation, or so poorly-functioning as to be obviously not fit for purpose. It's possible throwing more compute at the problem will work and they'll be useful in a few years. Or maybe we're actually experiencing a Markov-chain / ELIZA moment and are multiple decades away from anything actually useful.
I'll do you one further.
Self driving cars is far from a solved problem. It's something we've been working on for the last 20 years at least. We are getting closer, some solutions are impressive (like waymo). But even those ultimately need operators in the area of the cars to solve the problem of the car getting stuck.
Self driving cars are an infinitely simpler problem to solve vs a general purpose humanoid robot. You have basically 2 outputs, acceleration and steering. You have rules simple enough that I was driving at 14. With enough input, you'd think self driving could be completed in a snap. But it's not there yet.
Humanoid robots which are useful will come after widespread deployment of L5 self driving cars. Since we don't have that, I have no faith that we are close to useful humanoids.
However, self driving vacuums are a done deal.
"I have no faith that we are close to useful humanoids." I agree completely. We still have competitions about having robots walk up stairs and open doors, and balance while doing those things...
What would be the top uses of humanoid robots?
> What would be the top uses of humanoid robots?
To replace humans for general tasks. McDonalds would like to use them to flip burgers, take orders, and deliver meals. Amazon would like to use them to pick items from their warehouses for delivery.
I'd argue, though, in both cases the "humanoid" part isn't what they really want or need. Even in the mcdonalds case, so long as the robot is "cute" enough to interact with customers, then it can be a box with an arm.
Of course, a major problem with using robots in the food industry is cleaning and hygiene. Robots tend to have lots of hard to clean cracks that will happily carry around germs. Even in medicine, we are finding that things like laparoscopic devices are beasts that are almost impossible to fully clean.
They're already driving, as an example of why your claim of "gpt-2 moment" isn't as far out there as it may first seem.
https://www.1x.tech/
I have been following them since they were Halodi robotics. They're cool, but nowhere near the level of autonomy you need. My theory is that before household robots become a thing we should have self driving cars be a common occurrence, since that is a much much simpler problem.
We said the same thing about Waymo, that it was perpetually in the future. It took them less than a decade. The robots today are functionally capable, they don’t have the right fuzzy intelligence yet. It’s purely a data problem (lack of) and a lot of people are working on it.
Are you saying autonomous driving is a solved problem, even at scale? I haven't seen any Waymo in my small town in Southern Europe yet.
From what I've seen with Waymo, autonomous driving in relatively good conditions (light to medium rain, fog, or sunny) is a solved technical problem, even in small, winding streets. Snow is the next big hurdle, and they're actively working on that in Denver and Detroit.
They're being conservative with their rollout mostly because of civic issues. Most places do not have a legal framework for "what if your autonomous vehicle hits someone?" yet. Even if Waymos never were at fault for a collision with a person, you can always have cases like the one in Georgia back in October where a bicyclist wasn't looking where they were going and rammed into one. The shaky legal ground is a pretty big impediment right now, and that's in the US where we have much laxer laws about corporations killing people.
Yeah, I suspected that, but legislation and dealing with who is responsible and what happens if an autonomous vehicles directly kills/hurts someone or is involved in an accident it IS part of the problem, luckily.
Then, there will be concerns about the scaling costs. If it makes economical sense only in densely populated zones, what's the point? I mean, yes, it will be yet another business that works only in cities in that case, but that will not then make it universal as cars are.
If they can pull off year round driving in Manhattan, and they have, it’s a solved problem. Rollout it slow due to bureaucracy, appetite for progressiveness and catering to entrenched business interests but it will show up everywhere it’s profitable eventually.
> "small town" in "Southern Europe"
I've highlighted the two main issues you are currently experiencing.
Then it's not a solved problem.
It's not just a data problem, it's a hardware problem. Transformer-based robots require even more processing power than plain LLMs, as they also need to process visual and spatial/touch input. We don't have GPUs capable of fast inference on a SOTA LLM that would fit in a robot brain form factor, let alone also run fast enough spatial and visual processing. And there's currently nothing even approaching a feasible solution for cooling such a device.
If there's no unknown unknowns in the brain, it's most likely possible. As the universal approximation theorem and empirical results of scaling SGD+RL suggest. Whether it will be economically viable remains to be seen. The human cerebellum has a peculiar structure and 80% of the brain's neurons after all.
The parameter count equivalent of a human brain is not yet known, but if it was one per synapse then a full human brain replica would need about 1.5e14.
We also don't yet know how to be as efficient with training examples as any living creatures' brain, and we only partially make up for this by training on so many examples it would take you a million or so years to do the same, so we'd still stuggle with something proportionally smaller-brained such as a cat.
That said, remote controlled androids are going to be economically disruptive, as they make every (unlicensed) job open to outsourcing from an office in a low wage country.
Real neurons are orders of magnitude more complex than their artificial pseudo-approximation (it is all based on the century-old understanding of how neurons work). You can think of _individual_ biological neuron as an analog of the small artificial neural network. You can see this simple visual explanation on YouTube[1]. So we aren't even close. It doesn't mean the AI is impossible, it just means people underestimate the "computing power" of real brains, as well as that AI, even the future one might be totally different in how it works from the natural intelligence.
[1] https://www.youtube.com/watch?v=hmtQPrH-gC4
Brain's information flow also isn't just through direct neuron-to-neuron connections. Firing also releases neuromodulators into the extracellular space which affects how other neurons operate. Furhermore neuron connection architecture is very different between brain and feedforward ANNs, with the former exhibiting a lot of recurrent connections.
Deep learning doesn't try to mimic all the intricacies of biological processes. It tries to approximate the end result (information processing).
But for biological neurons to do something that can't be efficiently approximated on a digital computer (but conductive to useful information processing) they need to have unknown unknowns (well, partial unknowns like an unknown quantum algorithm will do too).
We don't know the violations of the physical Church-Turing thesis that are conductive for machine learning. We don't have evidence for their existence in the brain (although, the brain would be the prime candidate for finding them as evolution works directly with the true physical laws).
BTW, large ANNs don't try to model how the brain does things. They are trying to mimic what the brain does. So, using "how many transistors/artificial neurons it takes to model a biological neuron" is not a good approach.
We have no evidence. We even have no solid theories how this can work (Penrose's OrchOR is "OrchOR somehow taps into mathematical knowledge somehow encoded into the structure of spacetime"). But people, for some reason, insist that there should be something there. I can't attribute it to anything else but to deeply entrenched feeling of human exceptionalism.
You're talking about a philosophical debate whether the brain is computable, the other commenters are pointing out that even conservative estimates point to a brain-like NN requiring over a quadrillion parameters.
...assuming that modelling the physical structure of the brain is the only way to model its functions.
Building a "NN with similar capability as the brain" is not modelling its physical structure. The assumption is not made.
Let's define the terms. What does it mean "with similar capacity"? As far as I understand xvilka was taking about the number of artificial neurons required to model a biological neuron times the number of biological neurons in the human brain. It is modelling its physical structure (on a neuronal level).
You vastly overestimate how much we know about how our brains work.
Look up the neural correlates replication crisis, and e. g. the "dead salmon" study by Bennett et al.
That is we have unknowns that might increase or decrease estimates of computational demands of a functionally equivalent ANN. Not everything that happens in the brain contributes to information processing.
We certainly don’t need to replicate humans or overly anthropomorphise robots. Just like cars didn’t need to imitate mechanical horses.
If you're running a massive model for logic you're probably better off not putting it in the robot. And it'll be a long time before there's enough robots to make up a significant share of usage.
More basic movement control doesn't need loads of ram as far as I know.
The same logic for why self-driving cars can't be cloud based, applies for robots. Something cannot be in the middle of a delicate operation and then "oops!", no network, it just stops.
The larger the context window, the better with models. Having a few TB of RAM would be exceptionally helpful.
All this just made me realise something however. Having your robot dormant and charging, is a bit of a waste. You could have robots dormant, but its compute in use to act as a compute node. If the distribution of robots is similar world-wide, we'd need a fraction of the datacenters we have now.
Using such nodes for training purposes would be beyond advantageous. And the company which can slice up the work and having training done in batches would get the big bucks. And actually, with consumer facing products soon all laden with extra ram and gpu for local compute, that applies there too.
Imagine leasing out idle time on your desktop or even laptop for cash. There may be a market here, especially with the cost of new datacenters. Any company able to securely package compute without risking data safety is going to make a mint.
Anyone have any ideas?
> The same logic for why self-driving cars can't be cloud based, applies for robots. Something cannot be in the middle of a delicate operation and then "oops!", no network, it just stops.
I don't think you understood my post. The equivalent of self-driving is the movement control I was talking about.
Self-driving cars don't have high level logic, except for route planning. Which often is offloaded to the cloud. An extra 30 milliseconds on understanding your speech is nothing.
> Imagine leasing out idle time on your desktop or even laptop for cash.
https://vast.ai/
You're missing the point. The issue is "how is the thing used" not trying to make identical break points in differing tech. A self driving car cannot have object detection, collision avoidance, human detection offloaded to the cloud. Ever. At all. A network drop can't mean it smashes into things. Or stops unsafely.
The same is true with an android. Imagine it turning on an frying pan, cooking dinner, and then going offline part way through. Or turning on a tap to wash something, and going offline while the sink overflows and destroys the house.
There are myriad of such scenarios, but local compute is absolutely, 100% necessary. Anyone betting the farm on putting network controlled devices into homes for any serious task is going to lose their shirt. Local compute is an absolute requirement, and a few TB of RAM and local compute will be nothing over the scope of this discussion (a few years minimum, just to build and kick off all these new fabs).
By the time these fabs are online, expect most smart phones to have 1TB of RAM and significant llm capable compute (gpu or other custom silicon). I would be astonished if flagship model phones in 2030 weren't sold with 1TB RAM. Note I'm saying flagship, there will be of course economy models as always. Certainly laptops will be sold in multi-TB RAM configs.
I'm not missing the point. I'm saying that some amount of local compute is necessary but the really big stuff can be remote.
You don't need terabytes to turn things back off.
Not that I want to trust "not setting the house on fire" and "not flooding the house" to this kind of model in the first place...
> I would be astonished if flagship model phones in 2030 weren't sold with 1TB RAM.
I'll be astonished if they have 50GB.
Have you looked at RAM size/price trends? I'm not even talking about the last year, just the pattern before that. We're not in the 80s and 90s anymore. The most recent price lows were roughly $3.50/GB in 2013, $2.50/GB in 2016, and $1.50/GB in 2023. If we're lucky the cheapest stuff will hit $1/GB in a few years, and the kind that would actually fit on a phone motherboard would be significantly more than that.
Samsung hit 16GB on their top model in 2020 and it's either been 16GB or 12GB ever since. Apple only went up to 12GB in the last year. Google offers 16GB. A couple niche offerings have 24GB. Why would these RAM numbers even double during the next four years?
Well there's only one reason, so you do indeed know why.
Here's the thing, by 2028, maybe 2029, 1/2 the cloud AI providers are going to go bankrupt, or severely downsize. This will likely be due to massive reduced power and processor requirements for AI. GPUs are a stopgap, and they won't be used much longer. When that happens, and the number of servers drops by 90%, along with similar power requirements, a lot of people are going to go tits up due to massive, empty datacentres and compute.
That will absolutely flood the market with cheap obsolete GPU + RAM, used servers, you name it. Current RAM prices will plummet, and those able to withstand the storm will have to compete again, not just ride the wave like they've been doing now.
RAM has been ridiculously high for a long time, it won't be by 2030.
Couple that with local SoC with dedicated compute for LLMs, and you have the perfect storm for local, on phone compute. Why this? Well, because LLMs will be in everything! Certainly in robots, autonomous weapons, and anything the crazies can shove it into. Phones won't be the unique case, it'll be sharing a hardware market with robots, and 100 other things using the same SoCs.
Local compute will be easy... RAM makes a massive difference, huge as you know, difference in local compute. And with the datacentre market in tatters, and people becoming more and more concerned about privacy, people will likely not want to whisper to a cloud LLM, about their fetishes for donkeys or moose or whatever turns their crank.
Not to mention, other aspects of their lives.
In as RAM will be super cheap, that's where I see this going.
I have a sneaking suspicion you don't agree with this assessment. I guess we'll come back here 2030 and see what's what.
What do you think the production cost of RAM is? Even if we perfectly fixed competition and ended the AI bubble right now, I don't see how new RAM prices would drop below half the previous low point by 2030. Those fabs are legitimately ultra expensive, that's why almost all the RAM makers went out of business and we're left with the current cartel.
Also "massively reduced requirements for AI, so much so that we have 90% fewer servers despite the Jevons paradox" is not compatible with "high quality models make effective use of terabytes of memory" and "LLMs will be in everything".
And even if you could perfectly reuse all the memory out of the collapsing datacenters, that's not enough to make many of the phones and laptops you're describing. We'd need a 10x increase in manufacturing on top of the massive price drops, and even if we started building those fabs today they wouldn't be done in time (and we wouldn't have enough equipment to fill them).
So... like vertical markets? Ram for the data centers, data centers for the robots, robots for the Humanoid? or... the other way around? Robots to build data centers, and robots to build Ram.
I see... tech giants and R & D at the top of the pyramid, and I have seen Gumee, the fab that started all this Korean ram prodution.
[dead]
Humanoid robots that can do manual labor are going to be make or break for wealthy economies in the next two decades. Aging populations need help, and most successful nations do not have enough young people to do half the work they need done.
This is the path that Japan tried to go down and it hasn't worked out yet, but we have also solved a lot more of the technical issues since they began. going to be interesting to see if we pull it off this time.
Humanoid robots barely progressed between 2000-2020. There have obviously been incremental improvements in things like dexterity, vision, self-balancing, and locomotion, but in terms of having a useful humanoid robot, Honda's ASIMO released in the year 2000 is not crazily behind what we had in 2020. So it's not surprise we haven't seen economic dividends yet in the real world.
I think AI is what could make humanoids turn from parlor tricks to huge amounts of utility, but we're really going to have to see how it plays out in the next 5-10 years.
I think they mostly tried to go down this path before we had the transformer. With VLA models, or really now "Large Behavior Models", what's possible has changed dramatically. I've seen robot arms fold laundry now. Textile work is insanely hard, now it's just putting a lot of learned behavior together.
The current humanoid hype don't have much substances or key technologies in it, and incumbent industrial robotics companies like FANUC are already in the process of rolling the techniques created for humanoids into their robots. I personally think this is going to be just series of incremental gains for big welding bots, and nursery equipment becoming mildly robotic, like Aperture Science wall panels, than humanoids walking into retirement homes and doing dishes in the future.
Those were never techniques created for humanoid robots. Google researched using transformers as vision language action models to drive robots back in 2023 on a mobile manipulator and probably did non VLA work even earlier.
This is something people refuse to understand. The shape of the robot changes absolutely nothing about robot intelligence unless it abandons the basic concept of joints and links. Continuum robots are very difficult to control but they are also incredibly niche.
I really can't shrug off my suspicion since one of Chinese humanoid startups had added few female type body bumps on a robot and the online humanoid hype cluster("but does it do laundry or dishes" types) responded in panic with shared sentiments of defeat, that, those humanoid hype types might be just yearning for HBO Westworld bots, to state it with modesty, not even Data or Marvin who we can share drinks with. The absolute humanoid obsession doesn't make sense to me otherwise. The cookie cutter "humanoids are best poised for leveraging pre-existing human infra" response make close to zero sense otherwise.
Sounds like "That's a bold strategy, Cotton, let's see if it pays off for 'em."
I agree they may be strategically necessary. I’m just less convinced that “commercial humanoids by 2028” means anything close to replacing a meaningful amount of labor by then
Certainly. I would imagine that by the end of 2028, there are less than 1000 humanoid robots replacing a human worker. By the end of 29, maybe 5000.
>Aging populations need help,
They're pretty good at helping themselves. Close to where I used to live in Bavaria we had a pilot project of communal living for the elderly in a community of about a hundred people that included people with quite severe conditions such as dementia.
Medical and care personell routinely checked in but they were largely self sufficient and did a remarkably job of taking care of themselves, maybe most importantly the were happy and quite dignified, something I cannot imagine is the case when your only contact is a humanoid robot.
Of course in an age where every solution is yet another technology rethinking social life isn't very high up the list.
The "help" goes beyond taking care of themselves. What about food, clothing, infrastructure maintenance, and so on? An inverted population pyramid requires massive increase in the productivity of the economically active part of the population.
Or a regression in the standard of living. Ideally to a comfortable sweet spot for people.
The elderly already have a significant regression in standard of living, often catastrophically. That's a pretty big problem I would like to improve. There are so many people who don't leave their house because they struggle to get clean and they're embarrassed, and then their social connections collapse. It's going to be a long time before we have the technology to really help with elder care, but boy is it worth it.
It's also something we could help by getting rid of zoning in cities, of course, because we constantly push low income elderly out to the edges of urban areas where they can afford to live, rather than allowing supply to keep up with demand in dense, walkable, accessible places where people who can't drive can actually have quality of life.
I agree, we could do so much better. Zoning can be a tremendous quality of life issue, and needing a car to participate in society is an expense other people in other places aren't burdened with.
I honestly think the opposite.
Successful nations haven’t had enough young people to care for the elderly because they’ve all been employed as knowledge workers. Knowledge work is going away rapidly. The only work that will be left are things that require actual human bodies (for now).
In short, jobs like elder care are suddenly going to have an enormous number of new job seekers, since it requires human presence and isn’t taxing physical labor.
Are you familiar with Jevon's paradox? I think knowledge work is exploding.
Food has gotten progressively cheaper. At the same time, both the compensation demand for farm labor have plummeted.
Demand for the output of knowledge work can increase while demand for those who produce it falls. “Knowledge work” is not a product with inherent demand, just like you and I don’t go to the store to buy farm labor.
I'm not sure I agree with the premise of that analogy. Demand for farm labor has increased, but more of it is done by machines. The number of people employed in agriculture has increased if you include agriculture technology, over the same period. Not that I'm not saying the proportion of all humans working in agriculture has increased, I'm saying the raw number has, which is what matters to people along their lifetime.
Also, both the pay rates for agricultural labor and the pay for developing agricultural machinery are much higher than the pay for agricultural labor was in the past.
Knowledge work is an ever expanding umbrella, just as agricultural work is.
The title is kinda misleading. The actual wording is not specific about humanoid robots but "physical AI" which encompasses every machinery that can be potentially integrated with AI, especially focused on mass manufacturing for the Korean case. Basically this project is about all physical infrastructures to automate high tech manufacturing industry.
I think there's a reason they're being lumped together. Memory is the reliable industrial bet. Humanoids are the speculative "what will all this AI hardware actually enable?" bet
> Humanoid robots are like dance lessons, it is cool, it is sexy, and it may pay off in the future, but the value is much less certain
They need a solution to their plummeting birthrates which are officially worse than either China’s or Japan’s
https://en.wikipedia.org/wiki/Commodity:
> The wide availability of commodities typically leads to smaller profit margins and diminishes the importance of factors (such as brand name) other than price.
I'm not aware of many commodities which have only 3 world-wide sellers.
The value is pretty clear. The problem is the pay off is uncertain.
I’d say it’s more like “on groceries and a fancy dinner”. Humanoid robots sure do need RAM, both in data centres for training and in the robots themselves. :)
Vertical integration. Produce the chips, build data centers to run LLMs on the chips, and the robots to deliver the result to end users.
we will need humanoid robots since nobody makes kids anymore
I wonder how Germany missed the semi manufacturing train? They had literally everything: universities, manufacturing culture, expertise and supporting supply chains, cash.
I forgot, they also had ASML, freaking next door!
They had a large memory manufacturer, Infineon, who spun out their memory division as Qimonda which then went bankrupt [1]. They were the 2nd largest in the world at one time apparently. Looking back, it's easy to say the German govt should have thrown them a billion or two to keep them afloat. However, state intervention was very unpopular at the time in economic circles, and there was much furor over bailouts following the 2008 crisis.
Japan has an even sadder story. They were the DRAM top dog for a very long time. South Korea entirely ate their lunch.
[1] https://en.wikipedia.org/wiki/Qimonda
Even better, Qimonda was ultimately bought (alongside all their patents) by SMIC [1] who is now the Chinese memory player. For 30 million.
[1] https://newsletter.semianalysis.com/p/chinas-cxmt-is-set-to-...
It’s CXMT (memory) not SMIC (logic)
Wow, 7000 patents and all their IP and documentation
Infineon still exists as a semiconductor manufacturer. Their stock has gone crazy since start of the year as well.
Germany’s austerity policy after 2008 may be one of the largest economic blunders in history. It would be one thing if they merely committed self-harm, but they also used their pull in the EU to drag the rest of the continent down with them.
Reminded of Matt Yglesias’s excellent headline from 2010: Angela Merkel Lucky the Bar for “Worst German Leader” is Very High
One thing that I will never understand - how can german population not see through all that socialist bullshit she produced, promises undelivered, how much long term harm she generated across whole Europe (not just EU). Hard push for (now visibly) failed immigration at all costs, nuclear energy rollback (and she was nuclear physicist at least by her studies) and subsequent buying of foreign coal-based electricity, fragmentation of EU.
She literally licked putin's boots well into Ukraine war and still thinks licking his ass is the correct course to solve war in Ukraine (which started 2014 and it was pretty nasty already back then, the world just didn't care also thanks to her).
She is by far the biggest catastrophe modern Europe encountered after Hitler. She helped remove any proper fighting chance for the top dog Europe had for 21st century. She singlehandedly caused proper hate against EU in large parts of (not only) eastern EU population, and hence the rise of populist left or right wing politicians whose whole success story was just point at her failings and criticize, enough to get 20-30% of the votes and even win elections, repeatedly. She literally made people like Orban or Fico.
She still admits no mistakes, even wants to become german president. How effin' out of touch with reality she is.
> all that socialist bullshit she produced
... what? How's that the first thing that comes to mind about her, before "neoliberal", "conservative", or "austerity"? For that matter, when has the CDU ever been anywhere near socialist, in Germany or in the EU parliament?
100% agree we're still dealing with the fallout from her policies though.
okay, that is going too far
she has its flaws but remember that people vote for her, so its not only her fault
Yes people voted for her, she is still very popular. Doesn't change anything I said about her negative impact on entire continent that is extremely visible.
Maybe she did tons of good so it somehow averages out, but I certainly haven't heard about it (now is your time to defend her), everybody saw consequences of her disastrous policies that affected entire bloc.
its only hindsight, most people at those time didn't really foreseen the outcome later
This is untrue, entire eastern EU screamed like crazy against such immigration when she started importing immigrants en masse, "wir schaffen das". France faced the same with entire communities from Maghreb, most didn't work, 0 integration even after a generation or two spent in Europe, they knew language only if it was also their native one. Nobody in the east wanted that, entire political blocs before elections formed on opposing this and other fun EU moves (remember banning regular lightbulbs without good-enough LED replacements in entire EU market? Pepperidge Farm remembers... Or cutting down most of Borneo ancient rainforest to have effin' EU-subsidized bio fuels that ruin car engines quicker? Same story, I know this is EU but Germany is by far the strongest voice in and these are typical merkel moves - big moves regardless of consequences and picking up pieces later).
East had no prior experience with migrants due to living for decades in effectively prison camp guarded by soviets, no travel or other exposition to other ways of life. We were very monolithic cultures (and still mostly are).
The voices were completely ignored and overruled by behemoths like Britain, France and most powerful voice in the bloc - Germany.
Don't revision past, I lived through it, saw masses of people getting absolutely mad pissed off and feeling helpless and unheard with arguments which over time proved them mostly correct.
I don't hold those opinions myself, some of our best friends are muslim immigrants but oh boy go to the eastern EU without camera and ask random folks on the street outside capitals, or just listen to them. Or look at election results, this gave russians and their constant influence very good arguments since they position themselves as 'guardians of traditional family values in society', regardless of how its true or not (clue - its not but thats detail here).
You are racist
> Looking back, it's easy to say the German govt should have thrown them a billion or two to keep them afloat. However, state intervention was very unpopular at the time in economic circles, and there was much furor over bailouts following the 2008 crisis.
Absolutely, Germany essentially abandoned its position in industrial leadership solely due to neoliberal ideology. Just compare the trajectories of Germany and China in the last 20 years. One country planned and implemented a proper industrial policy, the other hummed and hawed about the infallibility of the market and thus essentially just gave up.
There were plenty of other DRAM manufacture before them as well. But the Boom and bust cycle wipe out those who weren't big enough or adequately prepared.
Nearly every single DRAM cycle we go through the same thing. People are saying memory are so cheap we are going to get 32GB/ 64GB / 128GB very soon. Just like they did 2-3 years before. Then every single DRAM up cycle they were earning insane amount of money, with this one being exceptionally high. Not only covering their previous lost but also extra profits. And every single time people cry out price fixing.
And every single time, Interviewer, financial reporters or share holders will ask TSMC's CEO Morris Chang the same question, and this is before TSMC was even well known in Tech cycle, will they make DRAM given it is just Fabs, TSMC could do it. And every single time in the past 20 - 30 years Morris had to explain the whole thing again.
Frankly speaking I have been rumpling about DRAM being commodity on HN for nearly 20 years. It is the first time I have seen a few comments finally picking this up as shown in the current top comment.
I don’t remember any time in recent history where demand skyrocketed so fast and so far away from supply. In previous cycles you mostly had supply driven shortages (remember all the fires?).
Maybe crypto surge was the closest to what we observe today? But again completely different magnitudes of demand. No company put 1T of capex for mining
> I wonder how Germany missed the semi manufacturing train?
My best guess is that the connecting train was operated by the Deutsche Bahn
lmao
Realistically, when it comes to the semiconductor market, there aren't many viable options outside of East Asia. I don't mean this in the sense that East Asians were somehow "chosen," but rather that the semiconductor industry inherently requires a large number of highly educated employees working together. The problem is that the working hours inevitably end up being very long. If you actually go work at one of those facilities, you have to wear a "cleanroom suit" (bunny suit), and it's physically demanding. What I'm saying is, you need highly educated personnel who can be mobilized at any time when a problem breaks out in the middle of the night, and who can be hired at relatively low cost. East Asia has a massive educational infrastructure — schools are very large-scale and the system is extremely well-developed — making it hard for other regions to compete. And indeed, the average working hours in countries that do semiconductor manufacturing are extremely long
In other words, it's an industry where you have to grind white-collar workers as if they were blue-collar laborers.
I’ve always wondered what is unique about semiconductors that PhDs need to work like assembly line workers. I’m sure they’re not solving partial differential equations all day, but what’s so different between different batches of chips?
The industry inherently deals with extremely hazardous chemicals, and on top of that, during semiconductor production, there are many things that have to be recorded and tracked.
A lot of the processes are automated, but at the points where automation hasn't reached, there are quite a few things that are genuinely complex to handle.
I think it’s more like highly skilled technicians, to scale up. Plus PhDs and other scientists to do the simulations and analyze the data for new designs.
Answer: they don't, they just work them into the ground because they can.
Mainland China also has the 996 schedule for office workers purely as a cargo cult ritual, forcing people to sit at a desk at midnight and pantomime doing work.
You're not wrong, but from what I understand, the issue really comes down to on-site personnel and supervisory staffing.
As for the 996 culture, I agree to some extent. My Chinese friends hated it too. But in China, there's this thing called neijuan (involution / the rolled up scroll). there are just so many job seekers that people are forced to endure it. What neijuan means here is "Eating one's own flesh" basically knowing that this competition is damaging to everyone involved but doing it anyway.
The people in the fabs aren't PhDs, they're extremely skilled technicians. They may have a similar number of hours of experience/training as a PhD, but that training is in troubleshooting automation systems, not doing research.
I believe some of the earliest Intel fabs were in New Mexico (Shiprock and Rio Rancho). What combination of the above did New Mexico have?
When New Mexico and Germany had fabs, South Korea was still a developing country ruled by a brutal dictatorship.
What happened was simple - both Taiwan and South Korea and now China took concerted steps in investing into their semiconductors businesses. South Korea did this indirectly through favourable arrangements for the industry players via the chaebol system, while China and Taiwan did this with more direct government investment into the industry.
Sure, you can't just dump money into the industry and become a semiconductor player, else the Middle Eastern countries would have tried that ages ago. Yes, the talent being locally present is important but you're once again bringing up tired tropes about Asian working culture as being relevant.
I'm not saying Asian culture is the main factor. Yes, it's true that authoritarian governance driven by dictatorial regimes and chaebol politics has played a strong role, but fundamentally, the long working hours are simply inherent to this business.
You brought up the New Mexico story quite well, but that place is notorious for the exploitation of Navajo women's labor. In the first place, the factory was occupied and shut down by the American Indian Movement. You know full well that this is a story about the exploitation of Native Americans, so why are you bringing it up like that?
The history of Shiprock itself is, at its core, a history of "cheap, obedient labor." You frame it only as state-led investment, but the reality is that the culture behind it is complex.
What my post is pointing out is not that "Asian culture is superior." What I'm pointing out is the harsh working conditions in Asia — where working hours are extremely long, and even highly educated workers are inevitably subjected to grueling hours. Why do you think TSMC's Arizona fab in the U.S. keeps getting delayed? The U.S. invested money through the CHIPS Act, but American engineers refuse to accept the "military-style 24/7 on-call readiness and brutal shift work" that exists in Taiwan. TSMC founder Morris Chang himself has pointed this out before.
What I'm saying is that the educational infrastructure is so well-established that it's easy to produce a large supply of highly educated workers, and that these highly educated workers then have to be submissive to inhumane working conditions. This isn't about Asian superiority — it's actually pointing out something bad about Asia. But from the context of your comment, it seems like you misunderstood me as saying "Asian work culture is superior" and replied based on that assumption. That was never my intention.
Before you leave a comment, I'd ask you to show some basic respect to others.
This is extrapolating from a single example of something that has worked and the conflating correlation with causation.
There are plenty of places with highly educated cheap workforces who work hard. Eastern European culture is almost identical down to the whole "tiger mom" stereotype.
And there are numerous counter examples: Ireland has a huge semiconductor industry: https://www.siliconrepublic.com/careers/semiconductor-compan...
The US is full of the "military-style 24/7 on-call readiness and brutal shift work" - at the high end silicon valley is built on this, and at the low end every single non-unionized factory is this.
TSMC has never built a fab outside Taiwan. Of course there will be problems.
There were probably many complex factors at play. Personally, I think the biggest one, as TSMC's Morris Chang said, is the inhumane working conditions imposed on highly educated workers. But there were likely also issues around permitting and regulatory procedures, as well as the overall cost structure.
As you said, if it were just about labor, other countries would probably have some supply of it as well. But in the case of Eastern Europe, there was likely American pushback against the European continent. As you know, semiconductors today can't be made entirely by a single entity. They're connected through a chain of trust. If Europe were to move beyond just producing semiconductor equipment and start directly manufacturing semiconductors through fabs, it would easily become a competitor to the U.S. rather than a supply chain partner.
In fact, the semiconductor chain is deliberately fragmented so that no single player can monopolize it.
On top of that, the U.S. is using South Korea and Taiwan to contain China. Under the ideology of protecting foundries from Chinese aggression and industrial attacks, the U.S. is sending the signal that it can cut off the supply chain. Eastern Europe, on the other hand, is tied up with the EU, making it much harder for the U.S. to control.
In the end, what matters when the consumer nation, the U.S., outsources production is how securely it can relocate it. Look at what happened to Japan's semiconductor industry. It was crushed through the 1986 agreement. The U.S. simply does not tolerate the emergence of an independent manufacturing hub that possesses sovereign economic power.
What matters is whether the U.S. can maintain control while keeping the price low.
Feels like the real clash happening here is that the reality is suggesting that the values of mean educational level of the bottom 99% of workforce outweighs that of the top 1%, and that being uncomfortable to some so much so that there has to be something else. But isn't that just it?
There's a story in one of Feynman's memoir where he figures out that pausing the live system and debugging its physical RAM stack is turning out to be more time consuming than simply scheduling a new corrected task, on some particular 1940s mechanical supercomputer he was assigned to as a tech. It might not have taken Feynman to notice that, but you can assign Feynman for that, and it worked for the Manhattan project.
The parent comment isn't (just) reiterating the tired tropes, but pointing out that East Asia has an "educational base" similar to industrial base that supports its high tech. I don't think that much is so strange way of thinking. The state of ME countries(maybe except Iran) soft proves it - they don't believe in such a thing. And they don't have a semiconductor industry. Pure coincidence? I doubt it.
(And on "This isn't about Asian superiority — it's actually pointing out something bad about Asia." from jdw64, yuuup 100% it is quad plus bad - IMO a thing about East Asia is that there's zero inter-national mobility due to the notoriously high language barrier, so competitions are closed to within borders, and the bar just drift skywards indefinitely because of that. There was a massive domestic hiring freeze in Japan during the 90s that made "janitors with a PhD" actually not so rare, but none of them hit the global labor market or started companies - the Japanese bar for janitors just went up to PhDs. It is said that success of Japanese 7-11 was partially attributable to that event, that, when you happen to have all the cashiers manned 24/7 with top scientists, you can just throw million different tasks and they can handle it perfectly, put aside whether they're happily doing it)
New Mexico has (pockets of) exceptionally well-educated population. 3 of the top 100 counties for PhDs per capita are in New Mexico (including #1, Los Alamos).
That's precisely my point. You need both the talent and the investment to execute this. Simply stating "Asian culture" is a stretch at best. The Asian economies made a ton of investment into semiconductors from the 1990s onwards, which are paying off now. Los Alamos is a pocket of exceptional PhDs because of the US government's investment into the institutional projects as well as building the infrastructure to attract talent to come and work in the middle of nowhere. That ended up creating a broader ecosystem of innovation there. What has the German government done, except for creating policies and regulations to ensure incumbents remain powerful? India pumps out more STEM graduates than most of the world combined, but what does it have to show for that in semiconductor technology but for a few middling projects?
Los Alamos would still be a desert without USG giving it a few billion dollars every year.
Chip fab locations have traditionally had more political than economic importance. Matrix multiplication chips and RAM have been the recent exception, while TSMC has long been the geopolitical exception. ASML's location only matters to the extent that it gets ordered not to sell to someone.
According to late Intel founder Andy Grove, having domestic manufacturing, including chip fabrication, is very important for a country's ability to innovate.
https://www.citizenstrade.org/ctc/wp-content/uploads/2011/05...
Not just ASML, but Zeiss the optics company is German and optics are one of the most important parts of lithography.
Regardless there are fabs in Germany, I know of at least 7, most located in Dresden. They mostly focus on older larger nodes, since for the German auto/industrial sectors that is all that is required. It's a much bigger industry than other European nations, especially like the UK which only really has one main fab in Newport.
The AMD spinoff GlobalFoundries has a fab in Dresden.
The whole semi manufacturing(/electronics) supply chain moved to Asia in the 1980s.
E.g. Siemens tried to compete but lost back then in the manufacturing part.
IIRC Taiwan took a page out of Singapore's playbook and went all in on electrical engineering and adjecent fields. It was very much a long-term strategy. Germany probably didn't feel nearly as much pressure, and was already very strong in all industry.
Intel was supposed to build a fab in Magdeburg, which would have been great, but apparently the reason it was canned (2025) was they couldn't secure enough customers.
apparently bosch semiconductor exists. whatever theyre doing over there they should do more of it tho, I want cheaper ram
Memory has only really recently become lucrative. Germany still has heavy machinery, trains, drilling machines etc all of which will be needed for a long time regardless of whether the "bubble pops" or not.
Most of those now need memory to function. At some point it becomes a national security issue.
That's not really a gotcha, because my train doesn't need a TB of dram.
Heavy machinery is starting to. Computer vision for robots is a big deal, and takes quite a bit of processing power. Robotic mining, earthmoving, and even construction equipment is exactly where Germany will innovate. Not to mention drones - Rheinmetall needs DRAM...
same thing that happened to Japan/US: they financialized their economy, companies moved from making things to pushing papers and outsourcing the making to China
Siemens?
> I wonder how Germany missed the semi manufacturing train?
Socialism or more its german variant, a system that can spend much more than more capitalistic power holders can ever earn and doesn't really plan well for future. Just look at it - very protected jobs, stifling bureaucracy, very hard to fire people, brilliant folks are definitely not compensated accordingly compared to (below) average peers - more often than not they earn the same. Its not agile economy nor workforce by any means, in contrary.
The feeling that the German Way (TM) is The Right Way, regardless of situation. If it worked in the past, it will in future, right. That leads to stagnation, complacency and when competition leap frogs them by the mile, surprised puzzled looks and wondering how it all happened.
[flagged]
Germany is done.
The Germany fetish still going strong I see.
people don't want to admit that German is fell off
> $585B on new fabs,
This is how overcapacity happens in a commodity. The is a rush to expand capacity to meet demand, and there will be overshoot. I hope these factories are built quickly so the memory crunch eases.
if you look at the price per gb of memory, at today's prices its only equal to 2014. People don't complain about memory prices back then. The reality is modern world just need much more memory. Before 8GB memory was enough, today my 32GB memory is 95% used all the time due to chrome. And my 2TB SSD is filled with 4K HDR videos. This is not counting massive storage needs by datasets, AI models, KV caches, and bandwidth needed for LLM inferencing.
it wouldn't be overcapacity in any shape or form
people literally cant give up AI right now and paid thousands of dollar for local AI model
5 years ago people would told you crazy spends 5k on such machine, now ????? people literally encourage you
Why humanoid? Surely there must be a superior physical form factor than one mimicking human anatomy. Is it just supposed to be more psychologically acceptable?
> Why humanoid? Surely there must be a superior physical form factor than one mimicking human anatomy.
There probably (certainly) is. But if you want to build a multi-purpose platform, you’ll soon be faced with a dumb challenge: nearly all interfaces (door knobs, taps, electric switches, cutlery, sponges, every single button out there, pillow cases, wrenches, hammers, signs…) are made for humans. Placed at human hand level. At human eye level.
Nearly all environments (houses, streets, sidewalks, factory floors, offices, toilets, bathtubs,…) are made for humans. Wide enough and tall enough (or short enough, for bathtubs) to accommodate human bodies.
So until we can find one or more form-factors superior enough to justify we adapt everything around it or them, betting that the easiest way to build a single multi-purpose platform able to do most things (and not n platforms for n+ use cases) is to borrow the shape most things are made for wouldn’t surprise me. Plus, you get a wider market.
And then, once you have happy-ish customers, figure out which of these human attributes and shapes aren’t actually needed to do the job.
Those require a hand on a stick. The stick isn't so interesting. Humanoid robots are the stick part, and actually not THAT interesting.
Never said it was interesting. Just that it’s easier to sell tech that fits your customers' environments than to get customers to overhaul their environment to fit the unproven tech you want to sell them.
Most fixed infrastructure like switches and buttons that can be operated with a single hand are amenable to any robot with at least one hand. We've had Sawyer demonstrations operating tools clearly designed for humans for more then a decade ago and Mobile Aloha had demonstrations showing how it can operate the switches of an elevator just fine. None of these are humanoid robots.
The moment you have mobile tools, whats the point in forcing the robot to hold them using a human hand? You can put them on a tool changer now or have a gripper that works for the specific task. Why does a robot need to hold a wrench using a humanoid hand?
>Nearly all environments (houses, streets, sidewalks, factory floors, offices, toilets, bathtubs,…) are made for humans. Wide enough and tall enough (or short enough, for bathtubs) to accommodate human bodies.
Uhm, now we're getting into stupid territory. All of those environments have flat floors. Flat floors are not an environment that are exclusively built to accommodate human bodies... The flat floor is designed for ultimate flexibility. It can be used for anything. Furniture, wheelchairs, wheeled robots, furniture on rolls, animals, and also humans.
All of the environments you've listed should preferrably be wheelchair accessible for disabled people (in terms of locomotion at least).
>So until we can find one or more form-factors superior enough to justify we adapt everything around it
Is this some kind of joke? Factories already make heavy use of UGVs and stationary robot arms and build custom end effectors for them. It's also an extreme strawman to suggest that wireless/electronic interfaces require finding a "superior form-factor" to the point that it feels insulting. There's also often an easy wheelchair accessible equivalent. E.g. a button to activate the electronic door opener at wheelchair level can still be at a comfortable height for standing people.
>And then, once you have happy-ish customers, figure out which of these human attributes and shapes aren’t actually needed to do the job.
So solve the impossible (come on you know it's hyperbole) first, only then can you build a simpler system.
I think its incredibly unreasonable to suggest that you need to solve every single problem ever encountered in human existence to be allowed to solve one much simpler problem.
I think we might be talking past each other.
I’m not arguing that humanoids are the only robots that can operate in human environments, nor that specialized robots shouldn’t exist. We already know they should, and do, and work very well.
I’m saying it depends on what you’re after.
If you want a specialized machine, then by all means optimize the form factor for that task.
If you want a general-purpose platform, then you’re faced with a near-infinite variety of environments, tools, and situations: homes, offices, hospitals, factories, hotels, streets…
Almost all of them were designed around human reach, dexterity, height, mobility and perception.
> The moment you have mobile tools, whats the point in forcing the robot to hold them using a human hand? You can put them on a tool changer now or have a gripper that works for the specific task. Why does a robot need to hold a wrench using a humanoid hand?
It doesn’t. The wrench was just one example among many. Nearly all our tools and interfaces are designed around human hands and bodies. If a different manipulator works just as well across all of them, great. My point isn’t that it must literally be a human hand.
> I think its incredibly unreasonable to suggest that you need to solve every single problem ever encountered in human existence to be allowed to solve one much simpler problem.
That’s not what I’m suggesting.
I’m saying it’s a deployment strategy. Start with the form factor that’s already compatible with the largest installed base of tools and infrastructure. Not because it’s mechanically optimal, but because it minimises the amount of adaptation required from the market.
And enables you to sell to markets you aren’t even aware exist.
If, over time, you discover that hands don’t need five fingers, legs don’t need knees, or an entirely different morphology delivers much better performance, then you have both the experience and the demonstrated value to justify changing the robot, or even convince your customers to change the environment to fit a better robot.
They’re basically trying to go for the "one size fits most" of robotics. And yes, we both know how well that usually fits anyone.
I’m not convinced either, but I can understand the logic.
There are just a few reasons - humanoid make sense, mostly for multi purpose tasks - where if you want a robot to be multi-job, do almost everything a human can do at work --
If you want a weld you need a 1 arm robot, if a robot to weld, then stack, then push parts on a cart across the factory - then sweep up, then etc.. etc.. perhaps a humanoid is alright.
There will definitely be too many people comfortable with ownership / master relationship with a humanoid robot that will do their bidding.
For multi-purpose tasks, human hands and arms are excellent, but only 2 are too few for many tasks. Humans very frequently need to have specialized gripping tools in order to accomplish tasks that cannot be done with only 2 arms, or they must work in pairs.
A good mobile multi-purpose robot should have 3 arms, or 4 arms for symmetry.
Human legs are normally not necessary. A mobile robot would just need some means to raise and lower its wheels, so that it could step when ascending or descending stairs.
A human head is not useful. The place for the "brain" of a robot is in its "chest", because robots do not have the limitations of living beings, where the very slow propagation speed of the nervous signals forces the nervous systems to be concentrated in the proximity of the main sense organs.
Instead of a head, one should have a couple of mobile arms with video cameras at their end, somewhat like the mobile stalks of crab eyes.
Of the components of a human, only the hands and arms are models useful to imitate. Cephalopod-like arms would be even more versatile than human-like arms, but it is likely that they would be much more expensive at similar performances.
Having the size of a human and human-like hands and arms is good for working in environments designed for humans, but having the shape of a human has no purpose.
I understand the argument but its honestly ridiculous in my eyes. How about a set of arms that can reach into dishwasher and stack dishes and a washer/dryer to fold laundry... Except even without solving the bipedal movement, that doesn't exist at a consumer price point.
Why are we pretending the hardest version of this is close to existing?
It doesn't need to be at a consumer price point first, it needs to replace a human at an existing warehouse or manufacturing role first, and that's achievable in the next two years at this point.
When you have arms that can reach into the dishwasher, you're also going to want them to put away your dishes. And so suddenly they need to get up high. And you're not going to have a SECOND set of arms at your washer/dryer to fold laundry, you're just going to buy a second DLC for your existing robot. And it needs to get between those places, so if you have stairs, wheels don't cut it. You need a bipedal robot very quickly.
Stair climbing systems that work using wheels exist. Google stair climbing wheelchairs for a few examples.
I am familiar, I'm a big fan of Dean Kamen's work. So far, we haven't seen a single wheeled stair climbing vacuum cleaner, even though the original iBOT is 23 years old.
That solves the horizontal mobility problem. And then you have cabinets - and wheels don't solve the vertical mobility problem. So then you need a scissor lift on those wheels, or a hydraulic lift.
The robotics nerds always end up back at bipedal because it's vastly simpler once you're already solving arms.
A scissor lift or some other thing seems a lot simpler than trying to allow for a bipedal robot to knee to reach things close to the ground. Or if legs are the best option, why not a quadruped? Much more stability with a lot of the same advantages.
I'm curious - have you done any robotics work?
Why not buy a second set of arms instead of legs or just a set a wheels?
I feel like if I write two paragraphs, nobody reads the second one...
Use wheels and buy two if you have to...the Roomba solution. Besides, why do you need to solve stairs the hardest way possible, a fully bipedal robot, before it moves past vapor?
Maybe they're asking what your argument against buying a second set of arms is, rather than suggesting it as a solution?
Totally, and they ask why not wheels...
I think the key is that none of our actual home use cases can be done with just arms. You don't need your folded clothes sitting in front of your washer and dryer, and a set of arms can't handle folding sheets.
Why not? I would love to have a set of arms that could flip the laundry from the washer to the dryer and then take it out of the dryer and fold it and put it in the basket.
I understand that particular use case sounds cool! I really do.
And then you want them to put away your dishes, and they can't, even though it's just a software update, because they're across the house. And they're BIG, so you don't have room to store two anyway.
And they were $20,000, so...
Back around the turn of the (20th) century, electric motors were expensive. It was not uncommon to buy one motor that could do multiple things, like this vacuum/grinder/buffer/blower/pulley:
https://www.youtube.com/watch?v=rw_8FWJuSho
If we start making robot arms at scale, they're going to get cheap.
I'm also not sure people are really going to want bipedal robots walking around their home, blocking the hallways, recording you in your underwear, etc.
Sure, an electric motor is like 10cm on a side. A set of robot arms that can fold laundry are like a 100cm cube. Most people aren't going to have space for two of them.
And the arms need cameras too...
A washer+dryer is pretty huge already. Seems like it could get some robot arms without changing the form factor dramatically.
That said, I think this is way farther off than anyone thinks. I want to know what the maintenance schedule looks like for robot arms. Looks like a lot of small moving parts. Probably a lot of plastic gears.
In an industrial setting, sure, maintenance is just an expense. But wheels require less maintenance and factories can be designed around the robots.
Plastic gears? I sure hope not!
As I said, maybe earlier in this thread, I've now seen a laundry folding robot work, but I think loading a dishwasher and then putting away dishes is going to come first.
Ok, so your idea is to sell a machine without software and then hand wave the software part away by saying it will be released at a later date? Sounds like a scam to me. You'll just end up with another AMD situation where the company is only interested in selling the hardware and has zero interest in developing the software, because it costs money and can only ever generate revenue by selling more hardware, which will make the hardware focused company feel vindicated in deciding to not put effort into the software.
My idea? No. I'm pointing out that if you release something capable of loading a dishwasher and putting away dishes, it could be as little as a future software improvement to make it do something else.
Human spaces are built for humans. Outdoors cars and quad coppers are a great form but constrained by stars, doors, and low ceiling makes them a poor fit.
Alternatively a 2 foot tall or a 20 foot tall humanoid robots aren’t particularly useful. But a good enough 5-6 foot tall humanoid robot can be swapped into an assembly line wherever a human is currently working without redesigning that workspace.
A lot of training data being collected is coming from people. You have companies paying people to do chores while recording themselves.
Backward compatibility with current meatspace tooling.
Tentacles & shoggothbots pave the way to the future!
It's just Internet hype.
Because you can use existing physical equipment with automation, until it’s ready for a full replacement
Because it's what Elon and China say that matters. There are exceptions but Korea is not the land of creativity. At all.
>The most costly of the megaprojects involves a commitment by Samsung and SK Hynix of $585 billion to build new chip fabrication plants..... The government’s goal is to double South Korea’s production of dynamic random-access memory (DRAM) within five years.
"Double"
I was wondering about that $585B number only to read they plan to Double the capacity. The two firm together is nearly 70% of the current DRAM market supply. I guess we are really going into an era where future Phone and PC will have 16GB as minimum, possibly more.
Even if the market are back to normal, that is doubling the revenue. The current AI scale and Cloud infrastructure buildout is simply beyond what I could comprehend.
> “Semiconductors, physical AI, and AI data centers are the triple axis for a great leap forward.”
Not the best wording... I wonder how serious this announcement is.
It's from the president's speech. Too lazy to look up the actual text but I guess he meant "pillars", a common metaphor in East Asia. In English axis and pillar are distinct but in East Asia the line is blurry.
For example, the Japanese word 軸 (jiku) is used to mean the "axis" of a graph, but it is also used in business to mean the "core pillar/backbone" of a strategy (e.g., 経営の軸 keiei no jiku, literally "the axis of management," but conceptually "the pillar of management").
The speech was delivered in Korean so this is a choice by a translator. I don’t speak Korean but I asked an LLM and it says …
the phrase used is "대도약" (daedoyak), which literally means "great leap forward" or "great jump forward." This is NOT "대약진" (daeyakjin), which would be the direct translation of China's "Great Leap Forward" (大跃进).
To expand a bit, even saying 대약진 _daeyakjin_ "great leap forward" wouldn't have turned many eyes, because _dae_ is just a common prefix ("great") and _yakjin_ is also a common word meaning "leap forward, push forward, improve". The word simply doesn't have the same connotation of the English phrase "Great Leap Forward", which is almost always used for the infamous Chinese movement.
If a Korean speaker wanted to talk about that Chinese movement, they'd use the full name, 대약진운동 (大跃进运动): the great leap forward movement.
Looks like a lazy translation; the president used a word "대도약" while the Chinese campaign that you're referring is translated into "대약진운동".
Top signal.
typical western main character syndrome
Better spend it now, people won’t need greater than 1.5tr parameter models
and battery powered consumer devices will be able to run those and lower sufficiently capable models by then, distributing the need for compute away from capital projects
the glut will be enormous
yes, immortalize this phrase just like the 640kb ram phrase, I’ll stand by it
> 1.5tr parameter models
Curious, what's this based off of?
utility that's good enough for a broad population
Why is the whole world jumping on to humanoid robots? What am I not seeing that requires this level of investment in it?
The great aging of working populations:
China: https://upload.wikimedia.org/wikipedia/commons/f/f8/Populati...
Japan: https://upload.wikimedia.org/wikipedia/commons/2/2e/Japan_po...
South Korea: https://upload.wikimedia.org/wikipedia/commons/4/4a/South_Ko...
United States: https://upload.wikimedia.org/wikipedia/commons/c/c9/US_Popul...
Europe: https://upload.wikimedia.org/wikipedia/commons/8/86/Europe_p...
Also the technology carries over to defense purposes
And then there’s the fact that tremendous investment is going into all things AI, and now hard tech
automating repetitive physical work doesn’t appeal to you?
Sure, but I'm not sure humanoid robots is the best form factor for this. E.g. something with a wheeled base makes movement calculations a lot easier since you don't need to deal with balance while moving.
But it needs more space to maneuver, a mostly flat ground, can't step over anything... It's simpler but less adaptable
For the optimistic case, read this piece by RethinkX:
https://www.rethinkx.com/blog/rethinkx/the-disruption-of-lab...
(Fwiw, >20 years ago RethinkX correctly projected the exponential cost declines of solar and batteries, when everybody else was drawing straight lines.)
> RethinkX correctly projected the exponential cost declines of solar and batteries, when everybody else was drawing straight lines
That’s not at all what I remember, what are you basing that claim on?
> Humanoid robots will enter the market at a cost-capability of under $10/hour for their labor, on a trajectory to under $1/hour before 2035 and under $0.10/hour before 2045.
I dont see it, unless this is an expectation that a robot will work for 50 years without maintenance at capex.
Why doesnt a comparable tool, like an excavator, work with this math? Why arent they 100 times cheaper to run than 20 years ago? Excavators can cost 50 - 100k pa in maintenance and fuel costs.
Why does creating a multifunction tool, with even finer tolerances, working in human safe workspaces cost less?
Because supposedly they would repair themselves or be repaired by other robots, and energy would cost less and less, anything would cost less and less if work is increasingly done by robots than can be improved year after year.
Thats a pretty huge assumption.
It reminds me of predictions for the cloud, that someone would just write a perfect cloud formation script and then be able to sack all of their employees. But instead it just updated a lot of System Engineers to Cloud Engineers and Devops positions.
The direct implication is that not only will robots definitely be allowed into human spaces, and definitely be able to use generalised hardware to perform fine detail technical work but also contain only servicable/replacable parts.
I dont see any of that as proven. I think its far more likely that a very limited subset of robots will be allowed to be near humans at all, and insurance for them is going to be extremely expensive. I think if I had a look at Australian law they would be considered Plant, guaranteed if they have any pressurized components or strong lift capability. Maintenance, is going to cost an arm and a leg, and the units themselves will be largely proprietary, so you are looking at ongoing maintenance via the vendor, much like what the tractor industry is working on.
But this logic applies to all machinery and technological progress. In the end it boils down to cheap energy and materials. The shape of the machine is actually irrelevant here.
Your "Fwiw" feels meaningless and hollow. The fact that there are stupid people out there, doing stupid things does not make you a savant just because you didn't make the same mistake.
This person clearly hasn't spent much time educating themselves how high volume manufacturing works, nor have they spent much time on flexible manufacturing systems either.
The entire article boils down to "Humanoids! Humanoids! HUMANOIDS!"
All current jobs have human input and output interfaces. If you want to sell new technology then it will be easiest to accommodate the already existing infrastructure.
The expected value is "infinite".
They can sell “employees” who don’t require salaries, onboarding, healthcare, 401k, benefits, etc and then leave after two years of being lazy and try to sue you. (This is how it will be marketed)
Economies like South Korea and Japan have a drastic population deficit that means there are simply not enough people around to perform many kinds of manual labor tasks.
Same as AI, replacing the workforce
Sex bots and disposable police. This is basically the future in every dystopian SciFi these politicians and oligarchs grew up watching. This is just living out fantasy.
why you acting like this isn't awesome ???
Cyberpunk is great game and I would like to have cyborg arm
What's the run-up time on a RAM fab? 5-10 years?
Depends. Still bottlenecked by ASML I would assume. Otherwise the current record holder TSMC manage to bring up a Fab operation and HVM in less than 18 months. Of course that requires a lot of planning before hand.
What is the big push for humanoid robots? We make non-humanoid robots because they're better at doing things than humanoid robots are.
>We make non-humanoid robots because they're better at doing things than humanoid robots are.
They're only better at some things, and some they can't do at all. If there's going to be a mass market for household robots, people are going to want a robot that can clean their toilet, read a book to their kids, unblock their kitchen sink, and climb a ladder onto the roof to get leaves out of their gutter, not a separate robot for each job. Any non-humanoid robot would struggle to do everything around the house that a human can do.
musk does it, so it must be good (TM), or something about cargo culting
I generally don't like the phrase cargo culting, but building a humanoid robot is probably the ultimate form of cargo culting.
The idea is this: You build a robot in the shape of a human with the hope that by building the robot well enough in the image of humans, it will become sentient and intelligent on its own.
That is literally the pitch of every single humanoid robot company on the planet.
South Korea is facing a serious demographic crisis, in the not too distant future it'll be a country of mostly elderly folk. I'd be interested to know if this investment has anything to do with this, since robots may be needed in the absence of young able bodied folk.
More info: https://www.youtube.com/watch?v=Ufmu1WD2TSk
It's crazy how much 1at could improve QOL for their citizens and also improve and diversify their economy. Alas, they're just going to subsidize ram prices for everyone when this current cycles goes from boom to bust.
Since this money belongs to Samsung and Hynix, it cannot be used for charitable activities. However, it is much better to build new cities, semiconductor factories, and power plants than to pay dividends to shareholders. The construction industry is one of the easiest ways to stimulate the economy.
What's incredible is how much resistance there is in the U.S. to do what for other countries is the obvious strategy forward. The U.S. — after decades of seeing manufacturing being outsourced — suddenly has an incredible advantage in data centers that is producing onshore facilities that are adding hundreds of billions of dollars in annual export revenue, and instead of there being a united front to maximize that advantage, there are huge obstacles being thrown in the way of the companies, administration and the state governments leading the data center expansion campaign, with Sanders and AOC calling for a national data center moratorium.
I live two miles (as the crow flies) from the site of a new construction of a data center. It’s being built by a European company. They will be installing gas turbines for power generation. Are we far enough away so that low frequency sound won’t affect our quality of life? We are seriously considering selling our home and moving. It’s a huge gamble for us. We are also noticing homes are staying on the market much longer than previously.
These data centers also get huge tax incentives because “jerbs”. Local politicians don’t understand after construction they don’t bring jobs or revenue to the city or county. This is where the resistance comes from.
That's a reasonable concern. Any industrial site is going to be a concern for nearby residences. But even proposed sites far away from any residential settlements are being opposed. The proposed National Data Center moratorium, for instance, is completely sweeping and would be devastating for U.S. strategic interests.
As for tax revenue, data centers contribute significantly to local communities.
https://www.wyedc.org/media/p/item/61886/data-centers-provid...
"Loudoun County, Virginia, often dubbed the "Data Center Capital of the World," provides a compelling example of how data centers can reshape a community's fiscal landscape. In 2018, the county hosted about 13 million square feet of permitted data centers. By 2024, that figure skyrocketed to 43 million square feet—a 231% increase in just six years.
This remarkable growth has substantially boosted Loudoun's tax base. The county's data center industry now contributes an estimated $890 million annually in tax revenue, nearly matching its entire operating budget of $940 million.
What makes data centers particularly advantageous is their cost-to-revenue ratio. For every dollar of tax revenue received from data centers, the county spends just $0.04 to support them, compared to $0.25 for traditional businesses. This financial efficiency has allowed Loudoun County to maintain the lowest real property tax rate in Northern Virginia—approximately 25% lower than neighboring counties."
this is what a country in decline looks like. When people oppose any sort of progress. Then will come acceptation or rationalization like the europeans do where they pretend they dont turn the AC to be enviromently friendly instead of just admitting they dont have the money for it
[dead]