Very interesting. It's rare to see people working on these non-mainstream models.
@nanotech_republika
Жыл бұрын
My question would be, is there in his model hippocampus and entorhinal cortex, etc. with its short term memory and other functionalities, like grid cells, place cells, etc.
@InquilineKea
Жыл бұрын
I first spent some time (maybe too much time) trying to learn a bit of this stuff in 2016, but LLMs came and I guess I don't feel guilty for not exploring this further given how little time there is in the world
@scientious
11 ай бұрын
Okay, let's go over this in more detail. 13:00 This model uses 6.5 million neurons to provide a functionality on the order of a fruit fly which has a brain with 2,500 neurons. Clearly, the model is lacking. 13:30 No, this a computational task rather than a cognitive behavior. It is entirely reactive. 15:20 Four elements? * Semantics ~ yes * Syntax ~ no, this would be computational at this level rather than cognitive * Control ~ no, this is also computational and motor control is irrelevant for a cognitive model * Learning & memory ~ these are required but this model's method is too limited 16:00 No, this is not biological cognition. This is a computational model that duplicates some of the lowest level brain functions. In other words, rather than providing a foundation that can be developed into something more sophisticated, this system is only a toy. It's like if you wanted to build a house and I gave you a truckload of bricks. You could stack them up to make a wall. You would find that without a foundation or mortar holding them together, you would have to make the walls thicker as you built higher or they would fall over. You couldn't make a door, windows, or a roof. You couldn't fix this by adding more bricks. 19:00 It is laughable to call these "executive systems". The model is not that sophisticated. 20:00 This functional diagram is definitely computational rather than cognitive. 21:49 No, this working memory model does not match human memory except at the neuronal level. It is missing almost all of the functionality of actual working memory. To be clear, there are fish with more complex working memory. This is not remotely at mammalian level, much less, human level. 25:33 Better character recognition is nice but it isn't cognitive. 25:48 RPM. ~ This one is particularly disturbing since RPM in a computational model relates to associative pattern matching, something that even slime molds are capable of and they don't even have neurons. In other words, rather than measuring fluid intelligence, this is only matching a learned pattern. You can tell this because if you vary the test slightly with a child, they will still get it right whereas Chris' model will fail because it doesn't have a matching pattern. Chris' understanding of his own work is somewhat delusional. 26:00 Vision is nice but this is irrelevant to a cognitive system. It's necessary for a reactive system. Again, Chris doesn't seem to understand the difference. 26:50 Motor control is not a cognitive function. 27:50 General Instructed Tasks. ~ Okay, let's see if there is anything more to this. 33:00 No, that's all computational. 34;00 No, this isn't a brain model except at a very low level like a fruit fly. None of the higher functions relate to brain functionality -- they are clearly computational. In other words, this is a computer model but much more inefficient. {sigh} I guess I can listen to the questions but the audience probably knows less about this subject than Chris, so I'm not hopeful. The Q&A discussion only confirmed that this is indeed only a computational model and has no relation to a cognitive model. Chris is like a proud parent watching their 3-year-old play with chess pieces and proclaiming that the child is a master chess player even though they actually know none of the rules or strategy of chess. Final thoughts. This video was posted only 7 months ago and Chris is a typical researcher. Still, it's somewhat shocking to me to see how very far behind the general research in this area actually is.
@scientious
Жыл бұрын
This isn't even close to accurate. This is not how to build a brain.
@selinawang5336
11 ай бұрын
Can you elaborate a bit please?
@scientious
11 ай бұрын
@@selinawang5336 Okay, I made a new post with more detail.
Пікірлер: 7