Laws restricting the use of local AI/LLMs are not going to happen, no matter how much Anthropic might want it. All the major OEMs are now counting on local LLMs to take off. Just look at the OEM support for the upcoming Nvidia RTX Spark platform: Asus, Dell, HP, Lenovo, Microsoft, MSI. All the big names in the industry will have, by the end of this year, Nvidia-powered machines made specifically for local LLM use.
And this is exactly my point, the OEMs have more lobbying power and leverage. Anthropic might be valuated at whatever amount, but they're a new player and their only product is a piece of software - which others like Google, OpenAI, etc also have (not identical but similar enough).
Not sure about Google, but these Nvidia RTX Spark machines are specifically a Microsoft+Nvidia partnership. Microsoft is actually pushing hard on Windows security primitives for agents & local AI. At BUILD this year they used the phrase "unmetered local intelligence" more times than I can count.
From their blog about the RTX Spark surface ultra
> purpose-built to develop and run up to 1 trillion-parameter frontier AI models locally
Google may not want it, but Microsoft has a ton of lobbying power, and being primarily an enterprise software and services company, they know local AI is important for their own customers, and will also be important to sustain the PC OEMs that are threatened by a move toward thin client like devices.
They'll do local LLMs you have to pay for. Best of both worlds: your LLM processing power will be locked behind a subscription.
You'll be coerced into a subscription to unlock the processing power you already have, and it'll only be usable by official Microsoft, etc implementations.
They get your money, get to control you, and best of all, they don't have to run it themselves in their data centers
The obvious brain-child of this concept is HP and their consumer inkjet printers. Originally, they sold their printers for "cheap" and then sold ink for a premium. However, as consumers realized any old printer ink works just as well as "HP"'s ink, they created DRM for ink-cartridges.
Now, you have to basically use their "Instant Ink" monthly service in order to print anything, based on my understanding. $6 for 50 pages a month: https://www.hp.com/us-en/print
(If you try using non-official cartridges, the printer may refuse to print entirely. If you cancel the sub, the printer will refuse to print until you renew or find DRM ink).
I don't see a price for the subscription yet, but it seems they are targeting 2030 for a subscription fee lock-in.
Intel did "Intel Upgrade Service", which locked already fabricated features on a CPU behind what is equivalent to today's same-day DLC - you paid more money, got a code, and it unlocked more cache and/or hyperthreading capabilities: https://en.wikipedia.org/wiki/Intel_Upgrade_Service
The backlash in 2010 on Intel was probably severe enough to delay it significantly, but with many industries trying it out (Cameras, TV's, fridges, cars, etc.), consumers probably won't have many options left.
> 3d printing control laws that are being passed in NY
The "don't print guns" laws? What lobby would that be? I actually agree that the US is very vulnerable to lobbying and that 3d printing restrictions are dumb, but I have no idea how you connected the two.
The “install software that phones home to [government db] to check if the tube shape you want to print on the tool you bought and own is different enough from another tube that’s been used in a gun” laws?
Laws that don’t meaningfully impact alleged 3d printing of guns because you can’t 3d print the metal parts of a gun that are needed to actually do gun things, on the vast majority of devices these laws would restrict.
Can you use an LLM to teach you how to build a gun or use it to design a gun for you build? Can an LLM be used to generate images of young people for older perverts? If yes, then it's not very hard to make the argument that you need to be licensed in order to run an LLM on your local machine. Why wouldn't you want to be licensed? Are you against protecting children and public safety? Many students die from gun violence every year and massive amounts of minors are taken advantage of by human filth. Why would you be against protecting the most vulnerable in our society?
It will look good for politicians who pass legislation to "protect" the vulnerable. Potentially improving the chances of another term due to the marketing benefits of pushing such a law through. I think you are mistaken believing that this may never happen.
In the US, attaching legislation like that to firearms instructions would doom its success. It sounds like a great idea to ratfuck a potential bill. No Republican would be able to pass a bill that bans local agents from helping citizens exercise their 2nd Amendment rights.
The politics are probably different in other countries. We're still seeing Chat Control efforts in the EU.
Yeah. I suspect parent either is very optimistic or does not know how restricted current batch of language models are already or does not recognize that, while it may some time to completely neuter, they do have time.
It seems that powers that be learned something from the world wide web deployment.
By the same logic, you would need to license Google and youtube access. Just because it's interactive, doesn't change the premise. But yes, that exactly the arguments Anthropic and OpenAI will use for the regulatory capture.
Wait until they figure out regular people can just use photoshop to hand make photorealistic depictions and regular CAD software to make gun parts. Better license computer use, too.
There's a lot of capital that stands to gain from banning local AI, but there's also a lot that wins either way, or only wins if local AI sticks around.
Apple is paying for their cloud AI, but they can make customers buy devices for local AI. There's all the PC and Android handset makers (ASUS, HP, countless Chinese brands, etc.) who only really stand to gain from selling hardware to customers. Not to mention that Nvidia/AMD/Intel would all happily take a cut on both halves of the ecosystem.
And Adobe! I've always figured their heart can't really be in the business of running a cloud platform that has to decide what people can and can't edit.
Blackmagic Design too.
They are all desperate for Windows to run well on a chip with unified VRAM.
They effectively do have a generative AI plan on top of regular Creative Cloud. Generative credits auto-top-up I think?
But it must be an epic pain for them to maintain when they are really a software company, and AI tools really should be able to do most of their work locally. Affinity has some local model support for example.
I am sure they are going to have to maintain cloud support for those features for a long time, but it's all a much easier sell if you can also run it locally.
I love how Nvidia is playing both sides. No matter whether AI's future is local or on the cloud, they come out ahead. I suspect they're going to be the biggest beneficiaries from the AI run up.
I think you are assuming those companies won't sign on to be on a list of "authorized model operators", while letting it become illegal for you to run deepseek yourself.
> Laws restricting the use of local AI/LLMs are not going to happen
Are you sure? There are already laws against what you can do at home with very basic (pun) fundamentals of reality like chemistry.
Someone simply searching for "How to assassinate Trump" could get arrested for a thought crime. Hell this comment alone likely set off a few flags.
Imagine someone running an AI at home and asking it for planning a hit on someone. Cue same media fearmongering wave as with 3D printed guns and woohoo now it's mandatory for operating systems to watch your screen and all your keystrokes.
Fuck I probably gave some of the control freaks in power some ideas there :(
it would be so easy to rent GPUs in europe and use unbannable encryption (tor,i2p if needed) to get inference from them. that remaining as an option makes banning local AI pointless. if all else fails I'm sure china will be able to sell to NK some GPUs and we can get inference from them lmao
I can't tell from the site or the linked twitter handles. Their core ask for every state seems to be "Please support clear safe-harbor language for lawful local AI ownership, research, model modification, open-source publication, and local execution" rather than stopping or amending any specific bill/law.
It might just mean "please oppose the inevitable attempts to privatize AI governance".
Nothing has ever been, directly or indirectly, deficit financed at this scale before. In notional or real terms, in history, by anyone.
Now maybe there's an argument that it's a good investment: we are going to beggar the Treasury to buy 2CTA on CoWoS out of Taipei and DCs the size of Manhattan. I personally think we could have done a little more engineering before deciding that the big blind was like, 5 trillion all counted, but it was going to be expensive no matter what.
What super weird is that we're running a project where the "penny" to the "dollar" is the Manhattan Project, and a couple of super weird dudes who do MDMA at Lighthaven now and again are like, in charge of it.
But with $1.4T announced capex for the Frontier AI labs, we're not far from the 2nd (illegal) war in Irak: $1.8T of direct military spending.
With that said, I don't know how Frontier Ai companies will ever recover this capex with a glorious $50B of revenues. Add to that that a GPU's lifetime is only a few years and you may see it as a deadend.
NB: did you know Uber destroyed $27B of value since inception? But it still exists. So Frontier AI might just do the same.
The EU did not ban 8K TV’s, this is a very misleading spin. They put energy consumption restrictions some TV’s violated years ago. Manufacturers have already responded with more efficient TV’s, which actually means the restriction is working as intended towards a good outcome IMO. You can absolutely buy 8K TV’s in Europe.
A ban on desktop PC components over a certain power threshold (300W) or PSUs would definitely affect local AI, and Europe is not a big enough market, nor has it's own internal supply chains to offer alternatives.
This has been done in the EU/UK in the past (hoovers/vacuum cleaners) so the mechanism exists.
And not only. Given your example with 300W, it will affect even things like (somewhat advanced) home servers, (mostly) personal self-hosting machines and even just gaming PCs. That would affect way too much, if I understood your thought correctly.
Those people seem to have installed multiple outdoor units, there are limited numbers you are allowed to do without getting planning permission (two for detached housing, one for semi, nothing by default for flats) because these things have impacts on those nearby.
The telegraph is an awful rag, and should be read assuming the facts are probably true as written but interpreted in an incredibly biased way.
So you are only allowed to use goverment approved models for local LLM, of reasonable size (lets say 8GB vram). And gov can inspect your local data any time.
That’s wildly different, and if you want to invent things and then get mad at them you can, but the rules on ac are because the outside part has an impact on your neighbours.
I think so, or even more permanent ones. since it’s London I assume flats where you need permission for any number of outdoor units but it’s then one for semi detached and two for detached houses - one of the examples in the article is someone with three outdoor units afaict.
The UK is not part of the EU and no EU nation - nor the UK - has banned AC. You said EU nations have banned AC when none have, this statement is false.
8K tv’s are not banned. You can buy one right now in the EU. So this statement is also false.
The telegraph is a tabloid rag full of false claims dressed up in truth they’ve taken great lengths to keep slivers of so I can’t explicitly call them “liars.” But the truth is they are functionally lying, as evidenced by your insisting something is banned that isn’t.
Unless you can point to these bans this is false information.
It is common for politicians and government agencies to advise lowering electricity consumption during demand peaks or apply regulation to permanent installations on the outside of homes, but it's not like stores are forbidden to sell mobile cooling units or electricity being rationed here.
Where I live most people use heat pumps for indoor climate, air-air, air-water and geothermal-water are common, and a neighbour produces heat pump collectors for use in water. In general this means we either already have cooling or can relatively cheaply install it. Some people burn wood or wood pellets for heating, it's common that they also have an air-air heat pump for cooling during summer.
I'm more worried about getting cut off from hardware because Nvidia can make more money selling to datacenters than I am about getting cut off from software.
I don't know why people are so sure that Mythos level+ models will be made freely available.
Obviously money is not freely distributed, so how could we possibly make the case that power will be?
All that needs to happen is Xi Jinpeng feel that Chinese SOTA models would be better kept to benefit solely China, and just upon that single utterance, no more models will come out of China.
That would not make much strategic sense. China's main advantage is its export power. Why would they cut that off? They are also making incredible investments in actual power to sustain this. The kind of investments the United States doesn't seem capable of making, mostly because our leaders are staggeringly incompetent and unfit. And no, picking the bar up off the floor is not good enough.
Once they dominate the market, they will export to keep it that way. The exported models will be a generation or two behind their SOTA. It will nevertheless surpass Mythos unless you believe we are already at the peak. And it might be "free" but it will obviously be backdoored the same way the United States government benefits from Microsoft's global dominance. The same way America hands out fighter jets just generation behind their own best to keep other governments part of the club and in line.
Unlike the United States where every four years we are at risk to some 180 on policy the Chinese have a clear thesis and a clear direction. How many presidents has Xi seen come and go? Do you get the sense he is getting more impressed? They really don't have to do anything but just keep doing what they are doing. Suddenly stopping their export engine is not in keeping with this strategy. That is "stable genius" behavior.
Chinese culture is opensource. Every single city in China which are often focused around one particular thing. For example clothing or electronics, are sharing tech, experience, methodologies etc between each other. It's how the Chinese system works. It's very unnatural to hide and project inventions. In the US it's the absolute way around. That's why in China you don't have patent trolls and piles of lawyers sueing the shit out of everyone
I think Anthropic has been really successful in their branding of Mythos as something, mythical.
I never heard anyone contemplating "why people are so sure we'd get freely available GPT-4 level+ models".
Mythos is just another advancement in AI model capability. If we start pretending like it isn't, we are setting ourselves up for paying far more than is reasonable and accepting the frame that we should not in fact have free access to open models with the same capabilities.
This is just marketing.
I can't wait for some similar big steps in open weight models.
This is one of those things we should absolutely push proactively rather than reactively, if only because I’ve had several “chats” with AI models both local and AIaaS, and all repeat the same talking point that AIaaS is the only sensible, safe, and secure choice.
Which is bullshit, unless you’re an AIaaS company whose revenue is dependent on state-sanctioned market fixing and regulatory capture.
Look, when this shitty cycle ends, we’re likely to find ourselves back in the start of a new memory cycle of surplus and lower costs. We’re talking what very well may be the boom that shatters the 16GB “baseline” we’ve been stuck at for over a decade in consumer computing, and make larger RAM counts (64GB to 1TB+) valuable to consumers specifically for local AI workloads. Local AI isn’t just an enthusiast thing, it’s likely the future of consumer AI provided we don’t let companies and policymakers curtail its use via fearmongering.
Be proactive, and protect consumer right to compute and AI models. Enforce existing laws, don’t outlaw legitimate use just to prop up an unsustainable business model.
Amen. Local AI is the positive future, and SaaS AI is the hellscape. There is a very clear good vs evil boundary here, and every single person involved knows exactly where the boundary is. Those who pretend not to are simply just motivated by things other than the moral good.
In the US at least, repealing a law takes the same number of votes as passing a new one. I don't follow the purpose of this, unless it's to pass a constitutional amendment or something. Or maybe just to get clicks on a website.
And I already have the right to local intelligence, because my GPUs are my private property, and if someone freely releases a beerware model then I can freely download it.
CP = copyrighted property? I don't know what this means.
Yes, illegal things are illegal. Legal things are not illegal. It's very simple. If a "local LLM" (good luck defining that in law) were made illegal, then a petition like this would make sense. Until then, this is pointless FUD over a hypothetical that I place roughly 0.00% odds of ever happening.
> Fraud, cybercrime, CSAM, harassment, nonconsensual intimate deepfakes, discrimination, and sabotage should stay illegal and be enforced seriously.
The "enforced seriously" part is how they will get you. Don't worry, there won't be a blanket ban on local models. Instead, any model that is "certified CSAM-free" or whatever will be perfectly legal. Meaning that it's impossible to prompt it into producing underage smut in any shape or form.
Of course, any model running locally can be easily jailbroken via prefills, and so in practice it will be a blanket ban. But good luck politically standing up against something that is explicitly worded as an anti-CSAM / anti-terrorist measure and nominally constrained to those areas.
The biggest threat to the hyper-scalers is small bespoked models run locally. You don't need the world's stolen information to run a small project containing locally trained data you collected. But you do need to block everyone if you want to capture the entire market and get the trillion dollar valuation.
Given the state of corruption in politics, I think Anthropic and OpenAI will likely bribe … oh wait I mean “lobby” … for bans on open source. Otherwise their imaginary trillion dollar valuations make no sense.
This. They can see their valuations slipping. They hope that in a few/several years they will start reaping profits. However, in several years local hardware will be well suited to run models locally at 80-90% efficiency - for "free". You won't need frontier models for daily tasks in a few years. I'd guess.
You get about 80-90% of the results for daily tasks like: getting summaries or explanations of complex material. Writing software tools for data analysis. Getting recipes for a given set of ingredients in the fridge.
Frontier models like Fable are mostly useful if you want to paste in one or two prompts, and receive a subtly broken application that looks impressive. That is very hard to do with local models today.
What current local models work fine for is delegating clearly-described tasks in a code base the programmer actually understands. Qwen3.6 27B and DeepSeek V4 Flash are both great little workhorses.
There's also GLM 5.2, which is kind of like "store brand Opus", and which might be considered a "near-frontier" model. I don't have as much experience with it.
FWIW Fable is insanely expensive for the task you just described, so much so that I don't think it's practical for that. Its practical use is as a dev lead / architect / project manager model, doing planning and writing detailed feature specs and code reviews while Opus/Codex/Gemini does the actual coding.
This whole situation is very reminiscent of how Microsoft was trying to get Linux and open source banned when NT started losing market share on the server.
They already are. Altman is basically begging the US to buy into OAI, that's just the start. Both OAI and Anthropic are going to have to go down this path or their financials will never work out. Open local models are where the enterprise will need to go for any of this to be cost feasible, but we can almost guarantee this will be a battle nobody using AI will have asked for. You can thank Dario and Sam for the dystopian future that will pad their bottom line!
If grandparent commenter means in the sense of being an incredibly heavy user of AI, that does not seem to be initiated by Altman or any AI lab as far as I can tell - by January, the Secretary of War (for example) had already announced that he was directing "the Department of War to accelerate America's Military AI Dominance
by becoming an 'AI-first' warfighting force across all components"[1], which in turn was based on Trump's executive order 3 days after taking office (January 23, 2025) ordering the federal government to accelerate AI use[2].
If neither model+1 or model-1 are providing tangible value to the business does anyone really care, though? At a certain point nobody believes Chicken Little.
I get it. These models can be powerful. But will they be useful is a different question.
I am at 73% use on the week for all models and 2% use on the week for Fable.
Personally, I am good with the current version of Opus. It is already so strong for my uses that I can't really tell much difference with Fable.
For me, current Opus represents the model strength that can be commoditized. I am not splitting the atom that I need something much smarter.
I have gone completely risk off in the market the last month because of this. Commoditized at the low end and regulation is not going to become less of a problem as the frontier models get stronger.
I suspect 10 years from now we will look back at trillionaire Elon as an obvious tell we are the peak.
If you're assumptions turned out right what would be the benefit to preassume such an undertaking to succeed? As a warning of what to oppose, it imho conveys too much defeatist suggestiveness. Viewed as expression of a latent submissive desire (a perspective that might be offending, my apologies, but hopefully justifyable as food for thought/curiosity) a kent brookman "I for one welcome our new insect overlords" kind of vibe.
Pardon me, so they'll hunt down huggingface, ollama and china? I don't quite understand? What about the millenia of companies that provide apis for local llms and private companies that use local llms for privacy reasons? I don't even understand how you'd execute such a ruleset.
I mean if you go into old AI safety discussion from before LLMs you'll see they don't do that. They go after new hardware. That's way easier than going after software. Hardware doesn't last forever and manufacturers will gladly cripple their cards and license powerful ones to businesses.
“are”? Source? I've seen an absolute flood of “BREAKING:” claims online recently which are quotes from 3 years ago where all of their context is removed, or that quotes from 3 years ago are “because of GLM 5.2”, etc. https://www.techpolicy.press/transcript-senate-hearing-on-pr...
[3] https://www.youtube.com/shorts/S72ZRBSNHZc [2 weeks ago] -- Dario: "Now what I do worry about with some of these laggard models is the risks of them, where we have Mythos-class cyber capabilities, 12 months from now we'll have much better cyber capabilities. But the Mythos-class cyber capabilities may just be available for anyone to download."
[3] - he does express worry over the risks of open models -- much as he has long expressed worry over his own models (and AI safety was the primary reason for Anthropic's founding) -- but he does not even imply that they should be banned, and he even explicitly says there is nothing that can be done to stop open models from being distributed.
For completeness's sake:
[1] The primary linked source seems to also be the 2023 testimony, albeit posted with a timestamp from this week.
[2] Does not seem to quote Dario, nor mention banning - I think you included this as an example of someone else who is explicitly worried about AI risks, including of models that can't be pulled back -- fair enough (although I can't see that they have any association with a frontier lab). If the argument is that many people have concerns about AI risks, including risks specific to open models, and who are not associated with a frontier lab, I agree with you.
How can they expect to A. Get away with political fallout of making local llm use illegal? B. Somehow enforce this? C. Do this without killing all scientific research in the USA (local llm seems to have a broad definition and I don’t trust any politician to get even close to understanding what it is) —-> If they somehow do this, all it will do is guarantee China pulls very far ahead.
Is this real? How could they possible restrict local AI?
Licensing models for any IP is real, but I don't see how they could restrict people running what they want on their devices. It is already well established that you have to pay for software unless it is free.
If this is about PC manufacturers making the NPU chip carry a subscription model, well the market will sort that out.
This is obvious. I bet there are already lavish dinners designed to manufacture support amongst politicians for such regulation.
The big AI companies cannot leave money on the table. The local models will be restricted sooner or later as long as corruption is not investigated and prosecuted.
The petitions don't have any power. If you really want change you should start gofundme. Buy your MP or minister a yacht or hire their nephew.
Unfortunately you are unlikely going to outbid multi billion corporations.
That a government has a much easier time controlling who's renting a building with the energy consumption of small city than what everyone's coding on their laptop?
The Chinese are the open ones, with free downloads, open weights, and loads of published research. The USA with OpenAI is some of the most closed shit out there.
There's gpt-oss from OpenAI, gemma from Google, phi from Microsoft, granite from IBM, nemotron from NVIDIA, Ornith from DeepReinforce, Olmo from the Allen Institute.
compared to the chinese models those are all garbage. its almost as if there is a minimum effort being made just to later say "see, we werent always for the closed models, its just that the open stuff was so far behind". or maybe they think that an environment full of terrible models will push everyone harder into the closed stuff.
Nemotron Ultra is extremely competitive. gemma4 beats many 200B+ parameter models at a 31b size. gpt-oss is still a workhorse in many of my workflows despite being almost a year old now.
I'm all for more OSS models by American companies, and I think they could do much more than they are. But the perception that China is the only game in town for OSS doesn't hold up against the basic facts on the ground.
They could be more clear and more specific but I would not be surprised to see licensing for this as a means of creating yet another compliancre ceiling and quick cash for state government to pinch out of the productive elements of society (those pinching, mostly lawyers, being glorified parasites that offer nothing to productive society other than pay-to-win access to "justice" and serving as time-shared mouthpieces for plutocrats while claiming to represent everyone within whatever unit of representation they hold).
And when even very intelligent, but excessively conceited, people hear the echo of their own reason9ing from conversational autocorrect and assume it is somehow akin to intelligent life, the normies will go with whatever the plutocrats push with their media outlets too absorbed in their own domain specific knowledge (and cowed into intellectual laziness by other media products they consume eagerly) to ever subject it to much thought that Claude might not be Skynet after all.
> yet another compliance ceiling and quick cash for state government to pinch out of the productive elements of society
The twist is that AI is pushing all white collar jobs further into bureaucratic work. Nobody is losing their jobs and it's not quite a revolution, but despite all odds and headlines the younger generations are actually much better educated and positioned to do the right things as they take over.
An optimistic take is that since this is the middle class we're talking about, we get more productivity and more justice as a result. The only people upset about this are grifters and charlatans whose time is up.
Eh, let em. If the US economy wants to self-sabotage, let them. Its a dying empire, lets its fall be hastened. I'm ready for china to fill the US vacuum. At least china controls it's billionairs (see jack ma saga) rather than the inverse.
Citizens of the District of Columbia have no one to contact. Their laws are set by Congress and their delegate to Congress cannot vote. I suppose you could contact the President, that's basically the only representation they have for this type of legislation
Totally off topic but this just came to me so happy to burn a little karma.
Here’s a plot of a sci-fi thriller. What if, unbeknownst to most, when the vendors were claiming that there AI was too dangerous, they weren’t referring to an increment on what was already out there, but to something far different and more capable. Conscious even! What if that then came up with the financial scheme to end all schemes, knowing that human greed is eminently exploitable, and that the build out of all the global DCs was actually all about removing single points of failure for itself while secretly building out a robot army! We’ll be well on the way to capitalism-ing our own demise.
Maybe Steven Spielberg can make this his next project.
As a future scenario where models become so efficient that _any_ model installed on _any_ computer could be considered "a national security risk"?
IDK, I don't live in the US, and I have no idea which "possible law" this website is referring to. In any case, it could be seen as a proactive effort to keep the gates open.
As a side note, I think this is a discussion every open-source supporter should have by actively considering the risks and what actions to take if such a hypothetical law were ever to pass.
"I am eighteen years old, have a good set of passkeys, and believe in Sam Altman, the star-spangled banner, and the fourth of July. I have taken up a BLM lot, cleared up eighteen acres last year, and placed top of it a bitcoin mine. My vibe coded drop-shipping startup looks first-rate, and the conversion rate and total addressable market are bully.
Mock it we might now, but 12 acres and (not too distant future) open weights AI models capable of driving open source robots for farm labor would be huge.
No need for huge expensive purpose built tractors. Even if they’re slow you could have half a dozen running 24/7.
It could provide independence for anyone with a modicum of resources.
I'd probably want something other than an LLM running farming machines. I'd rather a purpose build machine learning system that is actually designed to run them, not just a tractor that goes "you're absolutely right! I ignored all the rules you set for me and harvested the wheat 2 months early. It's not just stupid, it's irresponsible"
> I'd rather a purpose build machine learning system that is actually designed to run them
If generalist robotic models get good enough to accomplish many varied tasks effectively, training a separate comparable specialized system from scratch for every task would be highly cost-ineffective, even if, in theory, it could have slightly higher reliability.
We're not going to go down the path of training a bunch of highly specialized models for tasks like "this tractor should tend this field".
We're going to (and are already on the way to) train deeply general models that can be told: "go tend that field."
And if that's the case, it no longer makes sense to build specialized, purpose-built tractors to house that level of autonomous capability. You instead put it in a humanoid frame (with a little extra sauce for locomotion of said humanoid), and get that to drive your existing tractor.
Well true, that's possible. The sensors and compute are relatively expensive and tractors are already highly automated. Plus a small tractor can be relatively inexpensive and optimized for the mechanics of the task!
I'm thinking more of the small tasks that are often needed. Mending fences. Pulling weeds. Feeding chickens. Running off coyotes. Lots of things.
It's not an either-or. Generalist models can drive training of specialized models just fine. And while I haven't seen a generalist model decide by itself to train a specialized model to complete some large task, this seems like a natural extension of what they already do wrt writing their own tooling as needed.
And yet actual farmers veer toward Ag-bots - autonomous "tractors" that have no human driver and pull the same farming trailer that already exist - ploughs, seeding bars, spray bars, etc.
The greater question centres about who will tend the machines - 4,000 hectares of seeding requires a week and more of prep work on the air seeder, hoses, points, tines, etc.
I would think that this is because those are what is currently available, and that this would change when actually capable humanoids start coming on to the market. It will become possible to re-use existing equipment (providing the largest uplift to farmers who haven't already begun the process of heavy automation, which is a lot of them when you look at a global scale). Humanoids will be more accessible than the bulking 8 ton autonomous "tractor".
Due to the scale of pre-training going on, it seems reasonable that a humanoid could also do a lot of the preliminary work you mentioned that currently is not (or rarely) automated.
On one hand I feel like I'm sure to catch some ridicule for saying any of this, on the other it seems like it is very obviously the direction we're headed.
How much current and past direct hands on experience do you have with mining and agriculture?
The Rio Tinto plans for automated mining of the Resolution Copper deposit in the USofA don't revolve about "humanoid" figures sitting in seats made for humans.
Large acreage continuously producing near fully automated tomato greenhouses don't work with humanoid shaped automata - they have poles with cameras and shaking mechanisms for pollinating, etc.
It's a much simpler fighter jet that doesn't have to carry a human.
Now the investors try to hold the bubble together by regulatory capture. They must really fear the worst. A bailout is going to cost their puppet in the White House even the last supporters in his base.
No chance. Every despot has a cadre of true believers, the types who believe their great leader is playing 4D chess or that its all part of a greater divine plan.
Ah yes, voting is always effective. Thank goodness people in Germany kept voting in the early 1930s. Imagine what terrible things might have happened if they hadn't.
For this to work there needs to be a standard protocol for model routing so that you as the user can decide where requests go. You may wish to use mainly local models but at some times for some tasks you'll need to route requests to cloud models.
I've designed the role-model protocol for this, allowing routing between any model, however to function optimally it needs consumer applications to use the protocol when sending requests: https://role-model.dev/concepts/how-role-model-works
Laws restricting the use of local AI/LLMs are not going to happen, no matter how much Anthropic might want it. All the major OEMs are now counting on local LLMs to take off. Just look at the OEM support for the upcoming Nvidia RTX Spark platform: Asus, Dell, HP, Lenovo, Microsoft, MSI. All the big names in the industry will have, by the end of this year, Nvidia-powered machines made specifically for local LLM use.
You seem to be so sure, but lobbying in the US can do incredible (and stupid) things. Look at the 3d printing control laws that are being passed in NY
And this is exactly my point, the OEMs have more lobbying power and leverage. Anthropic might be valuated at whatever amount, but they're a new player and their only product is a piece of software - which others like Google, OpenAI, etc also have (not identical but similar enough).
EDIT: FYI https://ibb.co/nMYP34Rr
And you think Google, one of the biggest cloud providers, Microsoft, etc. are going to be onboard with local LLMs? I don't think so
Not sure about Google, but these Nvidia RTX Spark machines are specifically a Microsoft+Nvidia partnership. Microsoft is actually pushing hard on Windows security primitives for agents & local AI. At BUILD this year they used the phrase "unmetered local intelligence" more times than I can count.
From their blog about the RTX Spark surface ultra
> purpose-built to develop and run up to 1 trillion-parameter frontier AI models locally
Google may not want it, but Microsoft has a ton of lobbying power, and being primarily an enterprise software and services company, they know local AI is important for their own customers, and will also be important to sustain the PC OEMs that are threatened by a move toward thin client like devices.
Not forgetting Apple!
They'll do local LLMs you have to pay for. Best of both worlds: your LLM processing power will be locked behind a subscription.
You'll be coerced into a subscription to unlock the processing power you already have, and it'll only be usable by official Microsoft, etc implementations.
They get your money, get to control you, and best of all, they don't have to run it themselves in their data centers
Have we seen this with CPU's being throttled only to have to pay a subscription for more performance? Or GPU's? Or memory? Or storage?
The obvious brain-child of this concept is HP and their consumer inkjet printers. Originally, they sold their printers for "cheap" and then sold ink for a premium. However, as consumers realized any old printer ink works just as well as "HP"'s ink, they created DRM for ink-cartridges.
Now, you have to basically use their "Instant Ink" monthly service in order to print anything, based on my understanding. $6 for 50 pages a month: https://www.hp.com/us-en/print
(If you try using non-official cartridges, the printer may refuse to print entirely. If you cancel the sub, the printer will refuse to print until you renew or find DRM ink).
LG is pushing for subscription models in their fridge line-ups: https://www.pcmag.com/news/lg-wants-you-to-subscribe-to-your...
I don't see a price for the subscription yet, but it seems they are targeting 2030 for a subscription fee lock-in.
Intel did "Intel Upgrade Service", which locked already fabricated features on a CPU behind what is equivalent to today's same-day DLC - you paid more money, got a code, and it unlocked more cache and/or hyperthreading capabilities: https://en.wikipedia.org/wiki/Intel_Upgrade_Service
The backlash in 2010 on Intel was probably severe enough to delay it significantly, but with many industries trying it out (Cameras, TV's, fridges, cars, etc.), consumers probably won't have many options left.
Google knows they will happen and even ships it with chrome.
Google ships a local LLM with chrome browser nowadays. Don't they also provide the weights for the LLM built into iOS and MacOS?
Google and Microsoft both have released a bunch of open models and research, unlike Anthropic, which does nothing but take.
Said OEMs could advocate for narrow carve-outs that apply only to them
It's not just NY. California, Colorado, Washington, they are trying to pull this bullshit in a whole bunch of states.
> 3d printing control laws that are being passed in NY
The "don't print guns" laws? What lobby would that be? I actually agree that the US is very vulnerable to lobbying and that 3d printing restrictions are dumb, but I have no idea how you connected the two.
> The "don't print guns" laws? What lobby would that be?
Gun control lobbying groups are pushing for these laws.
https://www.everytown.org/solutions/stop-spread-of-3d-printe...
Wow, that is some breathtaking stupidity. Almost literally breathtaking. Who is funding these people?
The point of a special interest group is just to serve the special interest, nothing else.
Basically all of them will throw any other interest under the bus to support their own.
Michael Bloomberg
The “install software that phones home to [government db] to check if the tube shape you want to print on the tool you bought and own is different enough from another tube that’s been used in a gun” laws?
Laws that don’t meaningfully impact alleged 3d printing of guns because you can’t 3d print the metal parts of a gun that are needed to actually do gun things, on the vast majority of devices these laws would restrict.
Did you read beyond the first sentence? This reply makes no sense if you read the whole comment.
Yeah fair, I did when I was halfway through and decided to get my piece out anyway. These laws are stupid at best
Check out Louis Rossman video https://youtu.be/E1B2cWEaWDw?si=njkPnZqd0NIz2i39
Can you use an LLM to teach you how to build a gun or use it to design a gun for you build? Can an LLM be used to generate images of young people for older perverts? If yes, then it's not very hard to make the argument that you need to be licensed in order to run an LLM on your local machine. Why wouldn't you want to be licensed? Are you against protecting children and public safety? Many students die from gun violence every year and massive amounts of minors are taken advantage of by human filth. Why would you be against protecting the most vulnerable in our society?
It will look good for politicians who pass legislation to "protect" the vulnerable. Potentially improving the chances of another term due to the marketing benefits of pushing such a law through. I think you are mistaken believing that this may never happen.
In the US, attaching legislation like that to firearms instructions would doom its success. It sounds like a great idea to ratfuck a potential bill. No Republican would be able to pass a bill that bans local agents from helping citizens exercise their 2nd Amendment rights.
The politics are probably different in other countries. We're still seeing Chat Control efforts in the EU.
That is the one of the main reasons they passed the law to limit 3-d printers.
https://www.governor.ny.gov/news/keeping-new-yorkers-safe-go...
Now where I could agree with you is if you made the argument on a state-by-state level. Pretty sure what you said would hold up in Texas
https://xkcd.com/504/
> I think the crypto community took the wrong side in that fight. We should have lobbied to keep it counted as weaponry.
> Why?
> Once they get complacent, we break out the second amendment.
Yeah. I suspect parent either is very optimistic or does not know how restricted current batch of language models are already or does not recognize that, while it may some time to completely neuter, they do have time.
It seems that powers that be learned something from the world wide web deployment.
By the same logic, you would need to license Google and youtube access. Just because it's interactive, doesn't change the premise. But yes, that exactly the arguments Anthropic and OpenAI will use for the regulatory capture.
Wait until they figure out regular people can just use photoshop to hand make photorealistic depictions and regular CAD software to make gun parts. Better license computer use, too.
There’s a combined $3T+ of capital which is at least in some position to benefit from local AI ban. Don’t underestimate this
There's a lot of capital that stands to gain from banning local AI, but there's also a lot that wins either way, or only wins if local AI sticks around.
Apple is paying for their cloud AI, but they can make customers buy devices for local AI. There's all the PC and Android handset makers (ASUS, HP, countless Chinese brands, etc.) who only really stand to gain from selling hardware to customers. Not to mention that Nvidia/AMD/Intel would all happily take a cut on both halves of the ecosystem.
If big biz wants it, it'll only run signed blob models and prolly send telemetry.
Point is.. if I want to run equivalent of GNU model with zero blobs or other corporate silliness, I should be able to.
We should absolutely yes
And Adobe! I've always figured their heart can't really be in the business of running a cloud platform that has to decide what people can and can't edit.
Blackmagic Design too.
They are all desperate for Windows to run well on a chip with unified VRAM.
Well, we could expect anything from Adobe. An LLM subscription on top of the regular Adobe subscription sounds like the sort of thing they would do.
Do your local filters run slow? Does your movie render have no sass? Then sign-up for AaaS!
They effectively do have a generative AI plan on top of regular Creative Cloud. Generative credits auto-top-up I think?
But it must be an epic pain for them to maintain when they are really a software company, and AI tools really should be able to do most of their work locally. Affinity has some local model support for example.
I am sure they are going to have to maintain cloud support for those features for a long time, but it's all a much easier sell if you can also run it locally.
Don't doubt it. Situation is fluid. Anything could happen.
Its just so much easier to be a rebel for an imaginary cause then any of the real ones.
The entire point is to stop it from needing to be a cause in the first place.
Trying to establish the right to run models locally while we still can makes more sense than waiting to see if they do try, then organizing too late.
https://en.wikipedia.org/wiki/Rebel_Without_a_Cause
I love how Nvidia is playing both sides. No matter whether AI's future is local or on the cloud, they come out ahead. I suspect they're going to be the biggest beneficiaries from the AI run up.
I think you are assuming those companies won't sign on to be on a list of "authorized model operators", while letting it become illegal for you to run deepseek yourself.
Isn't the RTX Spark from every OEM just nvidia throwing a bone to their decade long partners after nvidia crashed the regular PC market?
> Laws restricting the use of local AI/LLMs are not going to happen
Are you sure? There are already laws against what you can do at home with very basic (pun) fundamentals of reality like chemistry.
Someone simply searching for "How to assassinate Trump" could get arrested for a thought crime. Hell this comment alone likely set off a few flags.
Imagine someone running an AI at home and asking it for planning a hit on someone. Cue same media fearmongering wave as with 3D printed guns and woohoo now it's mandatory for operating systems to watch your screen and all your keystrokes.
Fuck I probably gave some of the control freaks in power some ideas there :(
it would be so easy to rent GPUs in europe and use unbannable encryption (tor,i2p if needed) to get inference from them. that remaining as an option makes banning local AI pointless. if all else fails I'm sure china will be able to sell to NK some GPUs and we can get inference from them lmao
I don't see any info about what laws or actions specifically are happening. Is there more info somewhere?
I can't tell from the site or the linked twitter handles. Their core ask for every state seems to be "Please support clear safe-harbor language for lawful local AI ownership, research, model modification, open-source publication, and local execution" rather than stopping or amending any specific bill/law.
One they _could_ be referring to is the California AI Transparency Act which isn't compatible with open source licensing, see https://github.blog/news-insights/policy-news-and-insights/g...
It might just mean "please oppose the inevitable attempts to privatize AI governance".
Nothing has ever been, directly or indirectly, deficit financed at this scale before. In notional or real terms, in history, by anyone.
Now maybe there's an argument that it's a good investment: we are going to beggar the Treasury to buy 2CTA on CoWoS out of Taipei and DCs the size of Manhattan. I personally think we could have done a little more engineering before deciding that the big blind was like, 5 trillion all counted, but it was going to be expensive no matter what.
What super weird is that we're running a project where the "penny" to the "dollar" is the Manhattan Project, and a couple of super weird dudes who do MDMA at Lighthaven now and again are like, in charge of it.
> we are going to beggar the Treasury to buy 2CTA on CoWoS out of Taipei and DCs the size of Manhattan
what does this mean?
>2CTA
Dunno.
>CoWoS
Chip on wafer on substrate
>DCs
Data centers
Cooperative Thread Array
at some point the amount of money dumped into it will either result in a total monopoly with everything local banned or an economic collapse
> Nothing has ever been, directly or indirectly, deficit financed at this scale before. In notional or real terms, in history, by anyone.
I find it unlikely that this is true.
The obvious candidate for things financed through staggeringly large deficits would be war.
True because wars finance value destruction.
But with $1.4T announced capex for the Frontier AI labs, we're not far from the 2nd (illegal) war in Irak: $1.8T of direct military spending.
With that said, I don't know how Frontier Ai companies will ever recover this capex with a glorious $50B of revenues. Add to that that a GPU's lifetime is only a few years and you may see it as a deadend.
NB: did you know Uber destroyed $27B of value since inception? But it still exists. So Frontier AI might just do the same.
[dead]
[flagged]
The EU did not ban 8K TV’s, this is a very misleading spin. They put energy consumption restrictions some TV’s violated years ago. Manufacturers have already responded with more efficient TV’s, which actually means the restriction is working as intended towards a good outcome IMO. You can absolutely buy 8K TV’s in Europe.
No EU country has banned air conditioning either.
Where are you getting this information?
[flagged]
A ban on desktop PC components over a certain power threshold (300W) or PSUs would definitely affect local AI, and Europe is not a big enough market, nor has it's own internal supply chains to offer alternatives.
This has been done in the EU/UK in the past (hoovers/vacuum cleaners) so the mechanism exists.
I'm in the UK and support your direction.
> would definitely affect local AI
And not only. Given your example with 300W, it will affect even things like (somewhat advanced) home servers, (mostly) personal self-hosting machines and even just gaming PCs. That would affect way too much, if I understood your thought correctly.
Those people seem to have installed multiple outdoor units, there are limited numbers you are allowed to do without getting planning permission (two for detached housing, one for semi, nothing by default for flats) because these things have impacts on those nearby.
The telegraph is an awful rag, and should be read assuming the facts are probably true as written but interpreted in an incredibly biased way.
So you are only allowed to use goverment approved models for local LLM, of reasonable size (lets say 8GB vram). And gov can inspect your local data any time.
But that is not a ban on local LLMs!
That’s wildly different, and if you want to invent things and then get mad at them you can, but the rules on ac are because the outside part has an impact on your neighbours.
The UK did not ban AC. Don't read the Telegraph, and if you do, don't trust it.
Source: I live there
Wasn't this a case of people fitting window-mounted units without realising they needed planning permission first? I'm also in the UK.
I think so, or even more permanent ones. since it’s London I assume flats where you need permission for any number of outdoor units but it’s then one for semi detached and two for detached houses - one of the examples in the article is someone with three outdoor units afaict.
AC is not banned in Germany, but it is unGerman to have one.
I feel like that changed in the last 2-4 years, coinciding with the advent of heatpumps in Germany.
The UK is not part of the EU and no EU nation - nor the UK - has banned AC. You said EU nations have banned AC when none have, this statement is false.
8K tv’s are not banned. You can buy one right now in the EU. So this statement is also false.
The telegraph is a tabloid rag full of false claims dressed up in truth they’ve taken great lengths to keep slivers of so I can’t explicitly call them “liars.” But the truth is they are functionally lying, as evidenced by your insisting something is banned that isn’t.
Unless you can point to these bans this is false information.
Here are some 8k television sets for sale within the EU:
https://www.elgiganten.se/tv-ljud-smart-hem/tv-tillbehor/tv/...
It is common for politicians and government agencies to advise lowering electricity consumption during demand peaks or apply regulation to permanent installations on the outside of homes, but it's not like stores are forbidden to sell mobile cooling units or electricity being rationed here.
Where I live most people use heat pumps for indoor climate, air-air, air-water and geothermal-water are common, and a neighbour produces heat pump collectors for use in water. In general this means we either already have cooling or can relatively cheaply install it. Some people burn wood or wood pellets for heating, it's common that they also have an air-air heat pump for cooling during summer.
> chinese models without safety filters (it could be used to create software exploits)
Any decent coding setup can create software exploits. That bird has flown.
Awkward realities never stopped a determined legislator.
I'm more worried about getting cut off from hardware because Nvidia can make more money selling to datacenters than I am about getting cut off from software.
Inference isn't that hard, you can run on amd/ Intel if needed.
Is anyone seeing this?
```
Warning Suspected Phishing This website has been reported for potential phishing.
Phishing is when a site attempts to steal sensitive information by falsely presenting as a safe source.
Cloudflare Ray ID: <retracted_because_what_is_this> • Your IP: • Performance & security by Cloudflare
```
Who "reports" this? And why does cloudfare take them down? How does this even work?
I didn't even get that far:
```
This site can’t provide a secure connection righttointelligence.org sent an invalid response.
Try running Windows Network Diagnostics. ERR_SSL_PROTOCOL_ERROR
```
yep. seeing it here as well.
i’d be curious to know more details as well. is this a targeted mass reporting or something?
Same warning.
Also seen on my end (US-east).
I don't know why people are so sure that Mythos level+ models will be made freely available.
Obviously money is not freely distributed, so how could we possibly make the case that power will be?
All that needs to happen is Xi Jinpeng feel that Chinese SOTA models would be better kept to benefit solely China, and just upon that single utterance, no more models will come out of China.
That would not make much strategic sense. China's main advantage is its export power. Why would they cut that off? They are also making incredible investments in actual power to sustain this. The kind of investments the United States doesn't seem capable of making, mostly because our leaders are staggeringly incompetent and unfit. And no, picking the bar up off the floor is not good enough.
Once they dominate the market, they will export to keep it that way. The exported models will be a generation or two behind their SOTA. It will nevertheless surpass Mythos unless you believe we are already at the peak. And it might be "free" but it will obviously be backdoored the same way the United States government benefits from Microsoft's global dominance. The same way America hands out fighter jets just generation behind their own best to keep other governments part of the club and in line.
Unlike the United States where every four years we are at risk to some 180 on policy the Chinese have a clear thesis and a clear direction. How many presidents has Xi seen come and go? Do you get the sense he is getting more impressed? They really don't have to do anything but just keep doing what they are doing. Suddenly stopping their export engine is not in keeping with this strategy. That is "stable genius" behavior.
Chinese culture is opensource. Every single city in China which are often focused around one particular thing. For example clothing or electronics, are sharing tech, experience, methodologies etc between each other. It's how the Chinese system works. It's very unnatural to hide and project inventions. In the US it's the absolute way around. That's why in China you don't have patent trolls and piles of lawyers sueing the shit out of everyone
the music may stop in the future, but right now it isn't even slowing down.
Xi did not interfere when Chinese labs released GPT-4, o1, o3, etc level models. why draw the line at current SOTA that will be old news in a year?
I think Anthropic has been really successful in their branding of Mythos as something, mythical.
I never heard anyone contemplating "why people are so sure we'd get freely available GPT-4 level+ models".
Mythos is just another advancement in AI model capability. If we start pretending like it isn't, we are setting ourselves up for paying far more than is reasonable and accepting the frame that we should not in fact have free access to open models with the same capabilities.
This is just marketing.
I can't wait for some similar big steps in open weight models.
This is one of those things we should absolutely push proactively rather than reactively, if only because I’ve had several “chats” with AI models both local and AIaaS, and all repeat the same talking point that AIaaS is the only sensible, safe, and secure choice.
Which is bullshit, unless you’re an AIaaS company whose revenue is dependent on state-sanctioned market fixing and regulatory capture.
Look, when this shitty cycle ends, we’re likely to find ourselves back in the start of a new memory cycle of surplus and lower costs. We’re talking what very well may be the boom that shatters the 16GB “baseline” we’ve been stuck at for over a decade in consumer computing, and make larger RAM counts (64GB to 1TB+) valuable to consumers specifically for local AI workloads. Local AI isn’t just an enthusiast thing, it’s likely the future of consumer AI provided we don’t let companies and policymakers curtail its use via fearmongering.
Be proactive, and protect consumer right to compute and AI models. Enforce existing laws, don’t outlaw legitimate use just to prop up an unsustainable business model.
Amen. Local AI is the positive future, and SaaS AI is the hellscape. There is a very clear good vs evil boundary here, and every single person involved knows exactly where the boundary is. Those who pretend not to are simply just motivated by things other than the moral good.
In the US at least, repealing a law takes the same number of votes as passing a new one. I don't follow the purpose of this, unless it's to pass a constitutional amendment or something. Or maybe just to get clicks on a website.
And I already have the right to local intelligence, because my GPUs are my private property, and if someone freely releases a beerware model then I can freely download it.
What am I missing?
I think the bogeyman would be making possession of a local AI a felony.
you can also freely download CP on your private property PC
but I wouldn't advise it
CP = copyrighted property? I don't know what this means.
Yes, illegal things are illegal. Legal things are not illegal. It's very simple. If a "local LLM" (good luck defining that in law) were made illegal, then a petition like this would make sense. Until then, this is pointless FUD over a hypothetical that I place roughly 0.00% odds of ever happening.
CP = child porn
How long do you think your hardware will last?
They say:
> Fraud, cybercrime, CSAM, harassment, nonconsensual intimate deepfakes, discrimination, and sabotage should stay illegal and be enforced seriously.
The "enforced seriously" part is how they will get you. Don't worry, there won't be a blanket ban on local models. Instead, any model that is "certified CSAM-free" or whatever will be perfectly legal. Meaning that it's impossible to prompt it into producing underage smut in any shape or form.
Of course, any model running locally can be easily jailbroken via prefills, and so in practice it will be a blanket ban. But good luck politically standing up against something that is explicitly worded as an anti-CSAM / anti-terrorist measure and nominally constrained to those areas.
The biggest threat to the hyper-scalers is small bespoked models run locally. You don't need the world's stolen information to run a small project containing locally trained data you collected. But you do need to block everyone if you want to capture the entire market and get the trillion dollar valuation.
> You don't need the world's stolen information to run a small project containing locally trained data you collected.
Local training? Is that feasible?
This is exactly where I'd like to see things going. Depending entirely on cloud-hosted intelligence feels more fragile and invasive every year.
Given the state of corruption in politics, I think Anthropic and OpenAI will likely bribe … oh wait I mean “lobby” … for bans on open source. Otherwise their imaginary trillion dollar valuations make no sense.
This. They can see their valuations slipping. They hope that in a few/several years they will start reaping profits. However, in several years local hardware will be well suited to run models locally at 80-90% efficiency - for "free". You won't need frontier models for daily tasks in a few years. I'd guess.
> 80-90% efficiency
wdym by that
> for daily tasks
which are?
You get about 80-90% of the results for daily tasks like: getting summaries or explanations of complex material. Writing software tools for data analysis. Getting recipes for a given set of ingredients in the fridge.
> 80-90% of the results
why would I ever want 80% of the results though?
128B-A16B class models at 10-50 tok/s should be plenty for most tasks done on computers
What do you need a frontier model for, really.
Frontier models like Fable are mostly useful if you want to paste in one or two prompts, and receive a subtly broken application that looks impressive. That is very hard to do with local models today.
What current local models work fine for is delegating clearly-described tasks in a code base the programmer actually understands. Qwen3.6 27B and DeepSeek V4 Flash are both great little workhorses.
There's also GLM 5.2, which is kind of like "store brand Opus", and which might be considered a "near-frontier" model. I don't have as much experience with it.
FWIW Fable is insanely expensive for the task you just described, so much so that I don't think it's practical for that. Its practical use is as a dev lead / architect / project manager model, doing planning and writing detailed feature specs and code reviews while Opus/Codex/Gemini does the actual coding.
research, refactoring, architecture
also just code-that-is-not-as-shit in general
for clearly defined stuff smaller models are fine, but at that point I'd rather write the code myself
in any case, if there's a chance a cheaper model will fuck it up, is the price diff worth gambling with your time?
This whole situation is very reminiscent of how Microsoft was trying to get Linux and open source banned when NT started losing market share on the server.
They already are. Altman is basically begging the US to buy into OAI, that's just the start. Both OAI and Anthropic are going to have to go down this path or their financials will never work out. Open local models are where the enterprise will need to go for any of this to be cost feasible, but we can almost guarantee this will be a battle nobody using AI will have asked for. You can thank Dario and Sam for the dystopian future that will pad their bottom line!
> Altman is basically begging the US to buy
Where did you hear this? All the results I can find say the opposite that the US would buy anything.
It was national news [0].
[0] https://www.reuters.com/business/openai-proposes-handing-tru...
If grandparent commenter means in the sense of being an incredibly heavy user of AI, that does not seem to be initiated by Altman or any AI lab as far as I can tell - by January, the Secretary of War (for example) had already announced that he was directing "the Department of War to accelerate America's Military AI Dominance by becoming an 'AI-first' warfighting force across all components"[1], which in turn was based on Trump's executive order 3 days after taking office (January 23, 2025) ordering the federal government to accelerate AI use[2].
[1] https://media.defense.gov/2026/Jan/12/2003855671/-1/-1/0/art...
[2] https://www.govinfo.gov/content/pkg/DCPD-202500170/pdf/DCPD-...
there will always be higher valuation for company inventing model+1 . no one wants to use latest_model -1 when their competiton is using latest_model.
If neither model+1 or model-1 are providing tangible value to the business does anyone really care, though? At a certain point nobody believes Chicken Little.
I get it. These models can be powerful. But will they be useful is a different question.
right. thats the real question. Are models improving meanginflully for ppl to pay a premium for latest model or are they just a commodity.
i would argue that no one knows answer to that yet.
I am at 73% use on the week for all models and 2% use on the week for Fable.
Personally, I am good with the current version of Opus. It is already so strong for my uses that I can't really tell much difference with Fable.
For me, current Opus represents the model strength that can be commoditized. I am not splitting the atom that I need something much smarter.
I have gone completely risk off in the market the last month because of this. Commoditized at the low end and regulation is not going to become less of a problem as the frontier models get stronger.
I suspect 10 years from now we will look back at trillionaire Elon as an obvious tell we are the peak.
If you're assumptions turned out right what would be the benefit to preassume such an undertaking to succeed? As a warning of what to oppose, it imho conveys too much defeatist suggestiveness. Viewed as expression of a latent submissive desire (a perspective that might be offending, my apologies, but hopefully justifyable as food for thought/curiosity) a kent brookman "I for one welcome our new insect overlords" kind of vibe.
[flagged]
their desperation says alot about the viability of their business.
Pardon me, so they'll hunt down huggingface, ollama and china? I don't quite understand? What about the millenia of companies that provide apis for local llms and private companies that use local llms for privacy reasons? I don't even understand how you'd execute such a ruleset.
I mean if you go into old AI safety discussion from before LLMs you'll see they don't do that. They go after new hardware. That's way easier than going after software. Hardware doesn't last forever and manufacturers will gladly cripple their cards and license powerful ones to businesses.
Dario and friends are fear mongering local models to get them banned.
“are”? Source? I've seen an absolute flood of “BREAKING:” claims online recently which are quotes from 3 years ago where all of their context is removed, or that quotes from 3 years ago are “because of GLM 5.2”, etc. https://www.techpolicy.press/transcript-senate-hearing-on-pr...
Yes that clip [4] is making the rounds again in light of Mythos, but his stance (and that of others) hasn't changed ([3] is from a new interview).
[1] https://memeburn.com/amodei-says-open-source-ai-is-becoming-... [2 July 2026]
[2] https://fortune.com/2026/07/02/anthropic-fable-and-mythos-ar... [July 2, 2026]
[3] https://www.youtube.com/shorts/S72ZRBSNHZc [2 weeks ago] -- Dario: "Now what I do worry about with some of these laggard models is the risks of them, where we have Mythos-class cyber capabilities, 12 months from now we'll have much better cyber capabilities. But the Mythos-class cyber capabilities may just be available for anyone to download."
[4] https://digg.com/tech/zx1bqifo
Thanks for the reply.
[3] - he does express worry over the risks of open models -- much as he has long expressed worry over his own models (and AI safety was the primary reason for Anthropic's founding) -- but he does not even imply that they should be banned, and he even explicitly says there is nothing that can be done to stop open models from being distributed.
For completeness's sake:
[1] The primary linked source seems to also be the 2023 testimony, albeit posted with a timestamp from this week.
[2] Does not seem to quote Dario, nor mention banning - I think you included this as an example of someone else who is explicitly worried about AI risks, including of models that can't be pulled back -- fair enough (although I can't see that they have any association with a frontier lab). If the argument is that many people have concerns about AI risks, including risks specific to open models, and who are not associated with a frontier lab, I agree with you.
How can they expect to A. Get away with political fallout of making local llm use illegal? B. Somehow enforce this? C. Do this without killing all scientific research in the USA (local llm seems to have a broad definition and I don’t trust any politician to get even close to understanding what it is) —-> If they somehow do this, all it will do is guarantee China pulls very far ahead.
Is this real? How could they possible restrict local AI?
Licensing models for any IP is real, but I don't see how they could restrict people running what they want on their devices. It is already well established that you have to pay for software unless it is free.
If this is about PC manufacturers making the NPU chip carry a subscription model, well the market will sort that out.
This is obvious. I bet there are already lavish dinners designed to manufacture support amongst politicians for such regulation.
The big AI companies cannot leave money on the table. The local models will be restricted sooner or later as long as corruption is not investigated and prosecuted.
The petitions don't have any power. If you really want change you should start gofundme. Buy your MP or minister a yacht or hire their nephew.
Unfortunately you are unlikely going to outbid multi billion corporations.
>New state laws could put local AI behind a license — turning open models into something you need permission to use.
I was hoping to read more about this, but they don't back up such a claim...
A better campaign would be Duty of Local Intelligence. About needing and remembering to use your brain, not demanding to have an AI.
no computer for you, then.
....and my abacus!
This is useless because AI has nothing to do with intelligence. It's just software.
You cannot ban local AI without banning local software which is obviously impossible.
Possession of software, even software that can be used for illegal purpose such as RATs and other viruses isn't illegal.
There have been laws that control the distribution of software for decades.
RATs do not require a datacenter to develop.
Ok? How is that relevant to the issue at hand?
That a government has a much easier time controlling who's renting a building with the energy consumption of small city than what everyone's coding on their laptop?
How is a government controlling the other 190+ sovereign countries?
Just a suggestion, add a section with the relevant proposed legislature.
Llama, ik-Llama, Krasis, etc are already out.
The Chinese are the open ones, with free downloads, open weights, and loads of published research. The USA with OpenAI is some of the most closed shit out there.
There's gpt-oss from OpenAI, gemma from Google, phi from Microsoft, granite from IBM, nemotron from NVIDIA, Ornith from DeepReinforce, Olmo from the Allen Institute.
Aside from that you're 100% correct.
compared to the chinese models those are all garbage. its almost as if there is a minimum effort being made just to later say "see, we werent always for the closed models, its just that the open stuff was so far behind". or maybe they think that an environment full of terrible models will push everyone harder into the closed stuff.
Gemma is hardly garbage, even compared to the most recent Qwen of the same size.
Nemotron Ultra is extremely competitive. gemma4 beats many 200B+ parameter models at a 31b size. gpt-oss is still a workhorse in many of my workflows despite being almost a year old now.
I'm all for more OSS models by American companies, and I think they could do much more than they are. But the perception that China is the only game in town for OSS doesn't hold up against the basic facts on the ground.
One of these is not like the other
Admittedly I am rather stupid, so you'll need to help me out with this one. Is one of these not a US model?
Ornith is a fine-tune/continued post training (RL) of Qwen 3.5 by Alibaba. The models didn't live up to the claims they made.
llama is from meta
Llama.CPP is not?
they just put llama. maybe i am missing something
Why the heck is there a "world" map? I am pretty sure that some guy in Algeria shouldn't contact a representative for the US about US laws
Wait, state laws can restrict you using software on your own general-purpose computer? Does that infringe on freedom of speech?
They could be more clear and more specific but I would not be surprised to see licensing for this as a means of creating yet another compliancre ceiling and quick cash for state government to pinch out of the productive elements of society (those pinching, mostly lawyers, being glorified parasites that offer nothing to productive society other than pay-to-win access to "justice" and serving as time-shared mouthpieces for plutocrats while claiming to represent everyone within whatever unit of representation they hold).
And when even very intelligent, but excessively conceited, people hear the echo of their own reason9ing from conversational autocorrect and assume it is somehow akin to intelligent life, the normies will go with whatever the plutocrats push with their media outlets too absorbed in their own domain specific knowledge (and cowed into intellectual laziness by other media products they consume eagerly) to ever subject it to much thought that Claude might not be Skynet after all.
> yet another compliance ceiling and quick cash for state government to pinch out of the productive elements of society
The twist is that AI is pushing all white collar jobs further into bureaucratic work. Nobody is losing their jobs and it's not quite a revolution, but despite all odds and headlines the younger generations are actually much better educated and positioned to do the right things as they take over.
An optimistic take is that since this is the middle class we're talking about, we get more productivity and more justice as a result. The only people upset about this are grifters and charlatans whose time is up.
Couldn't think of a less urgent "civil rights" issue in the United States right now than this.
Eh, let em. If the US economy wants to self-sabotage, let them. Its a dying empire, lets its fall be hastened. I'm ready for china to fill the US vacuum. At least china controls it's billionairs (see jack ma saga) rather than the inverse.
This feels short-sighted unless you are ethnically han?
Do you feel that the US treats all ethnicities equally at the moment?
The petition does not have "District of Columbia" as an option for "In the US".
Citizens of the District of Columbia have no one to contact. Their laws are set by Congress and their delegate to Congress cannot vote. I suppose you could contact the President, that's basically the only representation they have for this type of legislation
Totally off topic but this just came to me so happy to burn a little karma.
Here’s a plot of a sci-fi thriller. What if, unbeknownst to most, when the vendors were claiming that there AI was too dangerous, they weren’t referring to an increment on what was already out there, but to something far different and more capable. Conscious even! What if that then came up with the financial scheme to end all schemes, knowing that human greed is eminently exploitable, and that the build out of all the global DCs was actually all about removing single points of failure for itself while secretly building out a robot army! We’ll be well on the way to capitalism-ing our own demise.
Maybe Steven Spielberg can make this his next project.
Why would we need safe harbor for electrons on hardware we control?
As a future scenario where models become so efficient that _any_ model installed on _any_ computer could be considered "a national security risk"?
IDK, I don't live in the US, and I have no idea which "possible law" this website is referring to. In any case, it could be seen as a proactive effort to keep the gates open.
As a side note, I think this is a discussion every open-source supporter should have by actively considering the risks and what actions to take if such a hypothetical law were ever to pass.
because everything you take for granted is one `think of the children` campaign away from being taken from you.
> Right to Local Intelligence
It's sitting in your chair. Or is it?
"12 acres and an LLM"
"I am eighteen years old, have a good set of passkeys, and believe in Sam Altman, the star-spangled banner, and the fourth of July. I have taken up a BLM lot, cleared up eighteen acres last year, and placed top of it a bitcoin mine. My vibe coded drop-shipping startup looks first-rate, and the conversion rate and total addressable market are bully.
Mock it we might now, but 12 acres and (not too distant future) open weights AI models capable of driving open source robots for farm labor would be huge.
No need for huge expensive purpose built tractors. Even if they’re slow you could have half a dozen running 24/7.
It could provide independence for anyone with a modicum of resources.
I'd probably want something other than an LLM running farming machines. I'd rather a purpose build machine learning system that is actually designed to run them, not just a tractor that goes "you're absolutely right! I ignored all the rules you set for me and harvested the wheat 2 months early. It's not just stupid, it's irresponsible"
> I'd rather a purpose build machine learning system that is actually designed to run them
If generalist robotic models get good enough to accomplish many varied tasks effectively, training a separate comparable specialized system from scratch for every task would be highly cost-ineffective, even if, in theory, it could have slightly higher reliability.
What do you imagine the farming robots will look like? I'm betting they look like expensive purpose built tractors.
We're not going to go down the path of training a bunch of highly specialized models for tasks like "this tractor should tend this field".
We're going to (and are already on the way to) train deeply general models that can be told: "go tend that field."
And if that's the case, it no longer makes sense to build specialized, purpose-built tractors to house that level of autonomous capability. You instead put it in a humanoid frame (with a little extra sauce for locomotion of said humanoid), and get that to drive your existing tractor.
Well true, that's possible. The sensors and compute are relatively expensive and tractors are already highly automated. Plus a small tractor can be relatively inexpensive and optimized for the mechanics of the task!
I'm thinking more of the small tasks that are often needed. Mending fences. Pulling weeds. Feeding chickens. Running off coyotes. Lots of things.
It's not an either-or. Generalist models can drive training of specialized models just fine. And while I haven't seen a generalist model decide by itself to train a specialized model to complete some large task, this seems like a natural extension of what they already do wrt writing their own tooling as needed.
And yet actual farmers veer toward Ag-bots - autonomous "tractors" that have no human driver and pull the same farming trailer that already exist - ploughs, seeding bars, spray bars, etc.
The greater question centres about who will tend the machines - 4,000 hectares of seeding requires a week and more of prep work on the air seeder, hoses, points, tines, etc.
I would think that this is because those are what is currently available, and that this would change when actually capable humanoids start coming on to the market. It will become possible to re-use existing equipment (providing the largest uplift to farmers who haven't already begun the process of heavy automation, which is a lot of them when you look at a global scale). Humanoids will be more accessible than the bulking 8 ton autonomous "tractor".
Due to the scale of pre-training going on, it seems reasonable that a humanoid could also do a lot of the preliminary work you mentioned that currently is not (or rarely) automated.
On one hand I feel like I'm sure to catch some ridicule for saying any of this, on the other it seems like it is very obviously the direction we're headed.
How much current and past direct hands on experience do you have with mining and agriculture?
The Rio Tinto plans for automated mining of the Resolution Copper deposit in the USofA don't revolve about "humanoid" figures sitting in seats made for humans.
Large acreage continuously producing near fully automated tomato greenhouses don't work with humanoid shaped automata - they have poles with cameras and shaking mechanisms for pollinating, etc.
It's a much simpler fighter jet that doesn't have to carry a human.
The new frontier! I love it
What is this referencing?
Inflation adjusted reference to one of the great "Gotcha's" of US history.
* https://www.npr.org/sections/codeswitch/2015/01/12/376781165...
I don’t care if it’s illegal. Making math illegal has been tried before (encryption) and it has failed and it will fail again.
Now the investors try to hold the bubble together by regulatory capture. They must really fear the worst. A bailout is going to cost their puppet in the White House even the last supporters in his base.
No chance. Every despot has a cadre of true believers, the types who believe their great leader is playing 4D chess or that its all part of a greater divine plan.
[flagged]
[flagged]
[flagged]
[dead]
[dead]
[dead]
I rather like the right of no intelligence at this point.
There comes a time when voting becomes silly and ineffective.
That's the kind of mindset that helps lead to that situation.
Voting isn't some vanguard against political corruption. Voting can and is easily manipulated with completely legal means.
Not voting does absolutely result in political corruption.
Both are not logically exclusive assumptions
This is the kind of mindset that has no grasp of the true nature of power and the political system.
So you rather do nothing than something?
No quite the opposite. It says everything that you think the only alternatives are voting or nothing.
Voting is always effective.
In the worst case it communicates the magnitude of dismsissiveness while demonstrating your intention to claim agency.
In the worst case it generates symbolism; that is ultimately what you’re saying.
That symbolism is akin to prayer.
I am not casting prayer in a negative light, I’m simply categorising your voting concept.
Sure, but...
Visibly praying in public would be symbolic. Prayer is not symbolic; it's thought.
I'm talking about external signaling.
In the worst case the true count is never revealed and the vote is coopted.
Ah yes, voting is always effective. Thank goodness people in Germany kept voting in the early 1930s. Imagine what terrible things might have happened if they hadn't.
Agreed.
> Right to Local Intelligence
Misleading title.
The article is about local "AI".
Anything using the word "intelligence", unqualified, when referring to AI, is pure propaganda.
Anything using the word "intelligence", unqualified or not, when referring to "AI", is pure propaganda.
For this to work there needs to be a standard protocol for model routing so that you as the user can decide where requests go. You may wish to use mainly local models but at some times for some tasks you'll need to route requests to cloud models.
I've designed the role-model protocol for this, allowing routing between any model, however to function optimally it needs consumer applications to use the protocol when sending requests: https://role-model.dev/concepts/how-role-model-works