Paper title: Disinhibitory signaling enables flexible coding of top-down information in cortical networks
(should be qualified as in-silico visual systems)
Method: replicate fMRI findings of visual abstraction using simple networks to model what's essential
Gist: in tasks 'Inhibitory neurons that suppress other inhibitory neurons seem to pass key information from the “thinking” part of the system to the “sensing” component of the system'
I've heard the same for motor control: it's not that the cortex aims for one action; it aims for a bunch, but most are inhibited. (You see this in chaotic movement when inhibition fails).
So it's not really "think and see" but "what you see when you're doing a task".
(There's some analogy in there wrt (AI) exuberance effacing selectivity in investment decisions...)
I'm very curious about inhibition failures in brain areas, especially between visual perception and motor control. I'm no neurologist but your brain seems to generate a lot of imaginary interpretation when sensing the visual field, but sometimes there are short circuit like failure that leak those potential imaginary futures with you current real self (leading to strange uncoordinated or overlapping motor control signals)
See I did what I saw when doing a task like grabbing for my Guinness to take a swig and ended up hitting my laptop with the bottle when I brought the bottle to my body.
"inhibition fails" reminded me of this passage from Fear and Loathing.
"Ah, devil ether. It makes you behave like the village drunkard in some early Irish novel... total loss of all basic motor skills. Blurred vision, no balance, numb tongue. The mind recoils in horror, unable to communicate with the spinal column."
It’s worth noting that in neuro speak, inhibition does not necessarily correlate with what would appear to be a “dis/inhibited” person; it’s referring to a specific process where signals are blocked from propagating, and because the brain is made of a complex web of inter-modulatory loops, this can show up in unintuitive ways
e.g. signals from the default mode network getting in the way of task-oriented behavior, which can result in people appearing “inhibited” where in actuality they’re failing to inhibit irrelevant internal signals and (errant bottom-up) attention to them (this is the case in ADHD).
Why are they using neural nets to model observed behavior (different parts activated) and then applying them to biological neurons that work completely differently? Real neurons communicate using precisely timed spikes and each neuron does a bunch of local computation as well.
There’s debate over how much timing actually matters vs the rate of firing. Some people do believe in precise spike timing but I would say the general consensus is that spiking rate is a better measure of the current state of a system. There’s significant noise (as far as we can tell) in a neuron’s timing and it’s best modeled with a poisson process so we tend to think of it as rate coding which we can at least hand wave as viewing an RNN with a ReLU approximating.
Generally you can take a geometric view of this where certain features in a stimulus covary with neural activations in the same way they will with RNN “activations” which is at the real core of why people model things this way. The general idea being a dot product in an RNN can tell you something about what features are relevant for a task and we can look for hints of the same information being encoded in neural data. Certainly not everyone is in agreement on this though.
The actual experiment is basically training a relatively large RNN (1000 units) to do a very simple copy-esque task. The weight values are constrained to be either positive or negative at the beginning of training. The RNN could probably be way smaller and still be made to solve this task.
It isn't even a spiking model.
It seems really hard to go from this experiment to "we've learned anything useful about how brains actually work."
Independent of the research itself, the article makes it seem as if neuroscientist are just discovering the deep recursion all the way back to V1. The idea that this was a one way stream of information processing was discarded a long time ago. Those back projections probably serve lots of functions, but we can be pretty sure they are there to let current context bias the weights for quicker recognition and reaction...e.g. if your context includes snakes, your visual system will attune to recognizing snakes even faster.
"Compare the brain to something like ChatGPT or a large language model. We can do far more, across far more situations, on a tiny fraction of the energy — and without being trained on the whole internet. The brain got there through evolution, through the redundancy built into its wiring. Our models are recurrent neural networks, which are quite different from the transformers behind today's large language models. The goal is to work out these principles one by one and use them to make AI leaner and more adaptive. This inhibition-on-inhibition motif is one of them."
Reverse engineering how algorithms in the brain work is a really promising path towards making genuine AI systems which would make the current crop of LLMs obsolete.
You're just doing sophistry here. There are patterns in the brain that form which allow it to model the world and make predictions about it. If we can understand the nature of these patterns and implement them in a different substrate, then we'd be able to do similar type of inference on a computer. That's what algorithms are.
Forming patterns when processing information doesn't mean an algorithm exists. The patterns are emergent properties of underlying processes which can't be described in a computational sense the way you want.
Otherwise, you'd have to argue all consciousness is basically just an algorithmic function. Are you prepared to go all the way with that?
The entire context for this discussion is that these patterns are being implemented in artificial neural networks. Also, a pattern by its very nature implies that it can, in fact, be expressed as an algorithm. There is zero basis for the suggestion that emergent properties of underlying processes can't be described in a computational sense.
Also, not sure how anybody could reasonably argue that consciousness could not be described by an algorithmic function. It sounds like you subscribe to idea of dualism which requires belief in supernatural processes outside the material realm. Are you prepared to go all the way with that?
Again, I'm not aware of any credible scientific basis for the position that consciousness is some sort of a field which is a product of quantum processes. The most plausible explanation is that it is encoded in the pattern of electrochemical firings of neurons.
Consciousness itself may just be nothing more than an exhaust fume resulting from the functioning of the brain rather than something profound to be explained. It may simply be a phenomenon of the system observing patterns within the world and modelling itself in the process as part of the simulation it is creating. This is basically the HOT view of consciousness, and it's really the only theory that at lest provides some explanatory power for the origin of the phenomenon.
Paper title: Disinhibitory signaling enables flexible coding of top-down information in cortical networks
(should be qualified as in-silico visual systems)
Method: replicate fMRI findings of visual abstraction using simple networks to model what's essential
Gist: in tasks 'Inhibitory neurons that suppress other inhibitory neurons seem to pass key information from the “thinking” part of the system to the “sensing” component of the system'
I've heard the same for motor control: it's not that the cortex aims for one action; it aims for a bunch, but most are inhibited. (You see this in chaotic movement when inhibition fails).
So it's not really "think and see" but "what you see when you're doing a task".
(There's some analogy in there wrt (AI) exuberance effacing selectivity in investment decisions...)
I'm very curious about inhibition failures in brain areas, especially between visual perception and motor control. I'm no neurologist but your brain seems to generate a lot of imaginary interpretation when sensing the visual field, but sometimes there are short circuit like failure that leak those potential imaginary futures with you current real self (leading to strange uncoordinated or overlapping motor control signals)
Imbalance of inhibitory and excitatory appears to be the definition/state of a seizure.
Apparently, this even goes both directions, as there are also inhibitory seizures, leading to temporary paralysis! [1]
[1] https://www.medlink.com/articles/inhibitory-motor-seizures
Thanks a lot
See I did what I saw when doing a task like grabbing for my Guinness to take a swig and ended up hitting my laptop with the bottle when I brought the bottle to my body.
Maybe I have a bit of inhibition going on there.
"inhibition fails" reminded me of this passage from Fear and Loathing.
"Ah, devil ether. It makes you behave like the village drunkard in some early Irish novel... total loss of all basic motor skills. Blurred vision, no balance, numb tongue. The mind recoils in horror, unable to communicate with the spinal column."
It’s worth noting that in neuro speak, inhibition does not necessarily correlate with what would appear to be a “dis/inhibited” person; it’s referring to a specific process where signals are blocked from propagating, and because the brain is made of a complex web of inter-modulatory loops, this can show up in unintuitive ways
e.g. signals from the default mode network getting in the way of task-oriented behavior, which can result in people appearing “inhibited” where in actuality they’re failing to inhibit irrelevant internal signals and (errant bottom-up) attention to them (this is the case in ADHD).
Think of it like a relay with a normally closed contact.
Why are they using neural nets to model observed behavior (different parts activated) and then applying them to biological neurons that work completely differently? Real neurons communicate using precisely timed spikes and each neuron does a bunch of local computation as well.
There’s debate over how much timing actually matters vs the rate of firing. Some people do believe in precise spike timing but I would say the general consensus is that spiking rate is a better measure of the current state of a system. There’s significant noise (as far as we can tell) in a neuron’s timing and it’s best modeled with a poisson process so we tend to think of it as rate coding which we can at least hand wave as viewing an RNN with a ReLU approximating.
Generally you can take a geometric view of this where certain features in a stimulus covary with neural activations in the same way they will with RNN “activations” which is at the real core of why people model things this way. The general idea being a dot product in an RNN can tell you something about what features are relevant for a task and we can look for hints of the same information being encoded in neural data. Certainly not everyone is in agreement on this though.
The actual experiment is basically training a relatively large RNN (1000 units) to do a very simple copy-esque task. The weight values are constrained to be either positive or negative at the beginning of training. The RNN could probably be way smaller and still be made to solve this task.
It isn't even a spiking model.
It seems really hard to go from this experiment to "we've learned anything useful about how brains actually work."
https://www.biorxiv.org/content/10.1101/2023.10.17.562828v2....
Independent of the research itself, the article makes it seem as if neuroscientist are just discovering the deep recursion all the way back to V1. The idea that this was a one way stream of information processing was discarded a long time ago. Those back projections probably serve lots of functions, but we can be pretty sure they are there to let current context bias the weights for quicker recognition and reaction...e.g. if your context includes snakes, your visual system will attune to recognizing snakes even faster.
That’s…quite the epic surname.
"Compare the brain to something like ChatGPT or a large language model. We can do far more, across far more situations, on a tiny fraction of the energy — and without being trained on the whole internet. The brain got there through evolution, through the redundancy built into its wiring. Our models are recurrent neural networks, which are quite different from the transformers behind today's large language models. The goal is to work out these principles one by one and use them to make AI leaner and more adaptive. This inhibition-on-inhibition motif is one of them."
Cool!
[dead]
[dead]
Reverse engineering how algorithms in the brain work is a really promising path towards making genuine AI systems which would make the current crop of LLMs obsolete.
It’s not algorithms.
The Church-Turing Thesis disagrees.
Ok but algorithms aren’t “real”, they are human interpretations.
You're just doing sophistry here. There are patterns in the brain that form which allow it to model the world and make predictions about it. If we can understand the nature of these patterns and implement them in a different substrate, then we'd be able to do similar type of inference on a computer. That's what algorithms are.
Forming patterns when processing information doesn't mean an algorithm exists. The patterns are emergent properties of underlying processes which can't be described in a computational sense the way you want.
Otherwise, you'd have to argue all consciousness is basically just an algorithmic function. Are you prepared to go all the way with that?
The entire context for this discussion is that these patterns are being implemented in artificial neural networks. Also, a pattern by its very nature implies that it can, in fact, be expressed as an algorithm. There is zero basis for the suggestion that emergent properties of underlying processes can't be described in a computational sense.
Also, not sure how anybody could reasonably argue that consciousness could not be described by an algorithmic function. It sounds like you subscribe to idea of dualism which requires belief in supernatural processes outside the material realm. Are you prepared to go all the way with that?
Maybe. I think consciousness is like a field and it could attach to hosts or systems once they are able to support quantum driven processes.
Again, I'm not aware of any credible scientific basis for the position that consciousness is some sort of a field which is a product of quantum processes. The most plausible explanation is that it is encoded in the pattern of electrochemical firings of neurons.
Consciousness itself may just be nothing more than an exhaust fume resulting from the functioning of the brain rather than something profound to be explained. It may simply be a phenomenon of the system observing patterns within the world and modelling itself in the process as part of the simulation it is creating. This is basically the HOT view of consciousness, and it's really the only theory that at lest provides some explanatory power for the origin of the phenomenon.
[flagged]