What I find puzzling is that many people seem to assume that AI can be generated using a specific uniform neurological structure, while the human brain is actually made of many different parts, some older, some newer, some more connected, some more isolated, some inhibiting others, some potentiating others, some mostly signalling with this neurotransmitter, some using that etc.

I would assume it might actually be easier to get there with a uniform neurological structure, as you don't have the "legacy" infrastructure left over from millions of years of evolution.

However, I think that the first AI humanity manages to build will be more or less a copy of a human mind and only later will we learn how to construct minds "from scratch". Akin to how a beginning programmer will often scrape together bits from various sources to build his/her first program and only later can make original work.

Question is, whether those older parts are best viewed as legacy infrastructure, or as ASICs - parts that are in their local optimization minimas for functions they perform, and better than a uniform architecture?

I mean, a lot of the "older" brain is control circuitry that keeps everything running and regulated without our conscious thought. Everything will have that -- at some level. It might just be actual control circuitry and not anything tied to the "brain" (e.g. a PSU in a computer), but the function would need to be performed, and if it's not part of the brain, the feedback we get (e.g. about stress or pleasure) might be lost?

I think that a lot is control circuitry, like the bits that regulate digestion or body temperature. I don't see any reason that for a superintelligence these could not be consciously controlled. The main reason we have so many unconscious processes and heuristics in our brains is to limit the total power consumption, as that was super important back in the days when food was scarce. If power consumption becomes less important, you could do more and better thinking.

It doesn't require a superintelligence. It's possible to gain a certain amount of control over various unconscious processes by means of mental-training.

The parts are best viewed in terms of what they do, which is known for a lot of them. Neuro-science is an established field:

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

> What I find puzzling is that many people seem to assume that AI can be generated using a specific uniform neurological structure

This is a basic result in computer science: universal Turing machines can simulate any other Turing machine. If you accept strong AI, then you already accept the likelihood that the brain is reproducible via a Turing machine.

Exactly. The brain is not homogeneous. It has different parts with different functions.

> I think it's unlikely that the human mind tackles Go in the same way.

This is very certainly true, which is what makes AlphaGo interesting to watch and study. The human mind, even one that has trained on Go for years on end, will still work with abstractions and ideas that do not relate to the game. AlphaGo and other computers lack this attribute, as any and all abstractions they may have learned relate entirely to the game.

Any ideas about the "human perception" of Go they may have gleaned from games that are included in the initial training dataset, I suspect have long been supplanted by novel notions gathered during the phase where the Neural Nets played against themselves. These phases are documented in the AlphaGo blog from Deepmind[1].

I suspect that we may reach "human level intelligence", but that this intelligence will not arise in the same way. That is to say, computers will at some point match us in most tests of intelligence, but the solutions they devise will be completely novel.

[1] https://blog.google/topics/machine-learning/alphago-machine-...