My opinion on Idea 1: Embodied cognition sounds reasonable to me. I found the example mentioned in the paper too abstract to be compelling, so here is a more concrete example of my own: When humans do geometry, they might use spacial reasoning in such an intense way so that their eyes, arms, hands, etc. are engaged. This means that the amount of computational power which is being used on geometry exceeds the amount of computational power the human brain has.
This claim is false. (as in, the probability that it is true is vanishingly close to zero, unless the human brain uses supernatural elements). All of the motor drivers except for the most primitive reflexes (certain spinal reflexes) are in the brain. You can say that for all practical purposes, 100% of the computational power the brain has is in the brain.
Moreover, most of the brain is inactive during any given task. So for your ‘geometry’ example, most of the brain isn’t being used on the task, this is a second way your claim is false.
I want to advance a true hypothesis. The reason “embodiment” is necessary for human-like intelligence is because a “body” allows the developing intelligence to perform experiments on the real world and obtain clean information. For example, if you wanted an AI to learn how simple objects behave—say a set of blocks a baby might play with. You could feed the AI thousands of hours of video taken from cameras, or give it control of a robotic system with manipulators with similar capabilities as hands. And build a software stack with rewards for ‘curiosity’ - putting the blocks into states not previously seen and observing the results.
I think the latter case, where the AI system is performing direct manipulation, gives the opportunity to learn from thousands of A:B comparisons. Where A is “hands performed a manipulation” and B is “hands did a different manipulation or nothing”. This allows the learning system, if structured to learn from this kind of data, the opportunity to learn causality well and to develop a general model for manipulation.
Moreover this kind of A:B data (this is the scientific method in a different form) is clean information. All of the variables are the same except for the manipulation performed by the ‘body’.
I posit that “tool using” AI, which I think are more indicative of human intelligence than merely babbling like GPT-3 does, requires the AI to either have a high fidelity simulation or actual connected robotics in order for the system to become good at using tools. (as well as obviously the right structure of internal software to be able to learn from the information)
Note, of course, unlike science fiction, the AI system need not have humanlink bodies to satisfy the requirement for ‘embodiment’. Nor does it need to be connected to them in realtime—offline or parallel learning is fine.
This claim is false. (as in, the probability that it is true is vanishingly close to zero, unless the human brain uses supernatural elements). All of the motor drivers except for the most primitive reflexes (certain spinal reflexes) are in the brain. You can say that for all practical purposes, 100% of the computational power the brain has is in the brain.
I agree with your intuition here, but this doesn’t really affect the validity of my counterargument. I should have stated more clearly that I was computing a rough upper bound. So saying something like, assuming embodied cognition is true, the non-brain parts of the body might add an extra 2, 10, or 100 times computing power. Even under the very generous assumption that they add 100 times computing power (which seems vanishingly unlikely), this still doesn’t mean that embodied cognition refutes the idea that simply scaling up a NN with sufficient compute won’t produce human-level cognition.
Yes and first of all, why are you even attempting to add “2x”. A reasonable argument would be “~1x”, as in, the total storage of all state outside the body is so small it can be neglected.
I mean...sure...but again, this does not affect the validity of my counterargument. Like I said, I’m using as strong as possible of a counterargument by saying that even if the non-brain parts of the body were to add 2-100x computing power, this would not restrict our ability to scale up NNs to get human-level cognition. Obviously this still holds if we replace “2-100x” with “1x”.
The advantage of “2-100x” is that it is extraordinarily charitable to the “embodied cognition” theory—if (and I consider this to be extremely low probability) embodied cognition does turn out to be highly true in some strong sense, then “2-100x” takes care of this in a way that “~1x” does not. And I may as well be extraordinarily charitable to the embodied cognition theory, since “Bitter lesson” type reasoning is independent of its veracity.
I took the original sentence to mean something like “we use things external to the brain to compute things too”, which is clearly true. Writing stuff down to work through a problem is clearly doing some computation outside of the brain, for example. The confusion comes from where you draw the line—if I’m just wiggling my fingers without holding a pen, does that still count as computing stuff outside the brain? Do you count the spinal cord as part of the brain? What about the peripheral nervous system? What about information that’s computed by the outside environment and presented to my eyes? I think it’s kind of an arbitrary line, but reading this charitably their statement can still be correct, I think.
(No response from me on the rest of your points, just wanted to back the author up a bit on this one.)
Writing stuff down to work through a problem is clearly doing some computation outside of the brain, for example.
I’m not sure that this is correct. While making the motions is needed to engage the process, the important processes are still happening inside of the brain- they just happen to be processes that are associated with and happen during handwriting, not when one is sitting idly and thinking
Yes but take this a step further. If you assume that each synapse is 4 bytes of information (a bit sparse it’s probably more than that), 86 billion neurons times 1000 synapses times 4 bytes = 344 terabytes.
How much information do you store when you have 3 fingers up when counting on your fingers? How much data can a page of handwritten notes hold?
You can probably neglect it. It doesn’t add any significant amount of compute to an AI system to give it perfect, multi-megabyte, working memory.
My opinion on Idea 1: Embodied cognition sounds reasonable to me. I found the example mentioned in the paper too abstract to be compelling, so here is a more concrete example of my own: When humans do geometry, they might use spacial reasoning in such an intense way so that their eyes, arms, hands, etc. are engaged. This means that the amount of computational power which is being used on geometry exceeds the amount of computational power the human brain has.
This claim is false. (as in, the probability that it is true is vanishingly close to zero, unless the human brain uses supernatural elements). All of the motor drivers except for the most primitive reflexes (certain spinal reflexes) are in the brain. You can say that for all practical purposes, 100% of the computational power the brain has is in the brain.
Moreover, most of the brain is inactive during any given task. So for your ‘geometry’ example, most of the brain isn’t being used on the task, this is a second way your claim is false.
I want to advance a true hypothesis. The reason “embodiment” is necessary for human-like intelligence is because a “body” allows the developing intelligence to perform experiments on the real world and obtain clean information. For example, if you wanted an AI to learn how simple objects behave—say a set of blocks a baby might play with. You could feed the AI thousands of hours of video taken from cameras, or give it control of a robotic system with manipulators with similar capabilities as hands. And build a software stack with rewards for ‘curiosity’ - putting the blocks into states not previously seen and observing the results.
I think the latter case, where the AI system is performing direct manipulation, gives the opportunity to learn from thousands of A:B comparisons. Where A is “hands performed a manipulation” and B is “hands did a different manipulation or nothing”. This allows the learning system, if structured to learn from this kind of data, the opportunity to learn causality well and to develop a general model for manipulation.
Moreover this kind of A:B data (this is the scientific method in a different form) is clean information. All of the variables are the same except for the manipulation performed by the ‘body’.
I posit that “tool using” AI, which I think are more indicative of human intelligence than merely babbling like GPT-3 does, requires the AI to either have a high fidelity simulation or actual connected robotics in order for the system to become good at using tools. (as well as obviously the right structure of internal software to be able to learn from the information)
Note, of course, unlike science fiction, the AI system need not have humanlink bodies to satisfy the requirement for ‘embodiment’. Nor does it need to be connected to them in realtime—offline or parallel learning is fine.
I agree with your intuition here, but this doesn’t really affect the validity of my counterargument. I should have stated more clearly that I was computing a rough upper bound. So saying something like, assuming embodied cognition is true, the non-brain parts of the body might add an extra 2, 10, or 100 times computing power. Even under the very generous assumption that they add 100 times computing power (which seems vanishingly unlikely), this still doesn’t mean that embodied cognition refutes the idea that simply scaling up a NN with sufficient compute won’t produce human-level cognition.
Yes and first of all, why are you even attempting to add “2x”. A reasonable argument would be “~1x”, as in, the total storage of all state outside the body is so small it can be neglected.
I mean...sure...but again, this does not affect the validity of my counterargument. Like I said, I’m using as strong as possible of a counterargument by saying that even if the non-brain parts of the body were to add 2-100x computing power, this would not restrict our ability to scale up NNs to get human-level cognition. Obviously this still holds if we replace “2-100x” with “1x”.
The advantage of “2-100x” is that it is extraordinarily charitable to the “embodied cognition” theory—if (and I consider this to be extremely low probability) embodied cognition does turn out to be highly true in some strong sense, then “2-100x” takes care of this in a way that “~1x” does not. And I may as well be extraordinarily charitable to the embodied cognition theory, since “Bitter lesson” type reasoning is independent of its veracity.
I took the original sentence to mean something like “we use things external to the brain to compute things too”, which is clearly true. Writing stuff down to work through a problem is clearly doing some computation outside of the brain, for example. The confusion comes from where you draw the line—if I’m just wiggling my fingers without holding a pen, does that still count as computing stuff outside the brain? Do you count the spinal cord as part of the brain? What about the peripheral nervous system? What about information that’s computed by the outside environment and presented to my eyes? I think it’s kind of an arbitrary line, but reading this charitably their statement can still be correct, I think.
(No response from me on the rest of your points, just wanted to back the author up a bit on this one.)
I’m not sure that this is correct. While making the motions is needed to engage the process, the important processes are still happening inside of the brain- they just happen to be processes that are associated with and happen during handwriting, not when one is sitting idly and thinking
Yes but take this a step further. If you assume that each synapse is 4 bytes of information (a bit sparse it’s probably more than that), 86 billion neurons times 1000 synapses times 4 bytes = 344 terabytes.
How much information do you store when you have 3 fingers up when counting on your fingers? How much data can a page of handwritten notes hold?
You can probably neglect it. It doesn’t add any significant amount of compute to an AI system to give it perfect, multi-megabyte, working memory.