[egg1]

At its core, arithmetic is a deterministic set of rules that can be implemented with logic gates. Computing is just taking that and scaling it up a billion times. What is intelligence? How do you implement intelligence if nobody can provide a consistent, clear definition of what it is?

[mdp2021]

Same thing: we create models about how to solve the problem, not biomimicry models about how natural entities solve the problem - these are not necessary. They are on a lower layer in the stack.

[egg1]

Except that doesn't make sense if you can't articulate what the problem space is. We know arithmetic inside and out, and we've understood how to make mechanical calculators centuries before the dawn of electronics. "Intelligence" on the other hand is a nebulously defined philosophical concept.

What I see with these attempts at AGI is VC-funded circuses throwing shit at the wall, hardly checking to see if it sticks, and then heaping more on top. Nobody can explain how exactly transformer models are the building blocks of intelligence or how building on top of it will lead to real intelligence.

[mdp2021]

The unclear sides of the implementation are just the usual normal hurdles on the path. There are no impacting doubts about the goal.

The point was, let me remind you, that we do not see any need for biomimicry: we did not need to simulate any brain to implement counting. Similarly, there is no need to simulate a brain to implement a reasoner (and the problem is well defined).

> these attempts at AGI

They are just occasional events in the whole history of the endeavour.

--

Edit:

> normal hurdles

...or, in other words: "yes, if we had the recipe, it would be trivial". Yet we normally manage without.