4 Comments

"The answer to that question matters enormously. If you think biological neural networks are doing something special — that they contain some secret sauce that computers lack — you’ll tend to have strong priors against AGI being around the corner."

Good point

Expand full comment

great post, thank you!

Expand full comment

You’re reaching for a known result here. If you believe some bunch of biological neurons can be described by a suitably well-behaved function (and how would the brain work if they couldn’t?) then you can prove definitively that you can model the function as well as you please with an artificial neural net.

https://en.wikipedia.org/wiki/Universal_approximation_theorem

Expand full comment

"The answer to that question matters enormously. If you think biological neural networks are doing something special — that they contain some secret sauce that computers lack — you’ll tend to have strong priors against AGI being around the corner."

The secret sauce might in the ability to

1) generate serial based logic from a massively parallel system (from Douglas R. Hosfstdter's GEB),

2) while regarding those algorithms as a good map for the parallel systems in our embodied brains

2.1) which we have insight into by way our stream of consciousness as mentioned

2.2) as well as other more confusing poetic ways which give insight

2.2.1) but we don't have insight into this process (of 2.2) like we do more with mechanical logic operations (see 1).

Actually all of that is stolen from Hofstadter when I think about it, but my… —the context has changed since I read it in the mid-80s as a teenager.

Note: this does not mean the secret sauce is in the poetry., nor some other gap-filler.

I do suspect a hint of the secret sauce lies in the way we can generate objective reality, at least the notion of it, the illusion of it, from our subjective experience, and test it, check it, use it i.e. the gap between the two (subject… —object) is solvable, but likely to remain incomplete (out of touch from reality). There is no organ for truth, possibly as it does not exist.

Consider telepathy, or telekinesis, if it was possible, or even existed, surely evolution would select for it and it would be "everywhere".

Obviously empathy is an attempt at telepathy, but is it akin to the way we get LLM to chat with us?

Also, Use-case scenarios (in evolution) just need to survive, not necessarily be understood by the survivors.

Expand full comment