Reverse Engineering the Human Brain to Achieve AI

The ongoing debate between PZ Myers and Ray Kurzweil about reverse engineering the human brain is fairly representative of the same debate that’s been going in futurist circles for quite some time now. And as the Myers/Kurzweil conversation attests, there is little consensus on the best way for us to achieve human-equivalent AI.

...

Complete entry

Posted by jim moore  on  09/08  at  07:56 PM

Hey George,
Computational functionalism may be what dooms whole brain emulation. Just because a turing machine can in principle compute any physically computable function doesn't mean that it can do it on a similar time scale.

For example, consider the chemical reaction of a caffeine molecule binding to a receptor. These two molecules are performing a quantum mechanical computation to solve the Schrödinger equation for all of their particles. This simple system is finding the simultaneous solution for about 2^1000 equations. That is a task of such immense complexity that if all of the matter of the universe was recast into BlueGene supercomputers, they could not find the solution even if they crunched away for the entire history of the universe.

Computational functionalism may not allow switching computaional substraits without paying a houge cost in terms of time and energy needed to do the "same" calculation.

Posted by Tim Tyler  on  09/09  at  03:16 PM

I'm with Ben on this - as I think most people should be.

Posted by Mark Gubrud  on  09/13  at  05:28 PM

The Ben vs. Ray debate is a false dichotomy. AGI is most likely to be realized by whatever combination of technologies turns out to work best. I would predict a hybrid of digital and neuromorphic hardware, statistical learning, genetic algorithms, and rule-based programming. It won't be "reverse-engineering" the brain, but it will use many principles learned from neuroscience and study of the human brain, and also from linguistics, psychology, etc. It won't be one brute-force method or another, if for no other reason than because it is a complex many-faceted problem and some methods will work better for some parts and other methods for for other parts.

Posted by Paul  on  09/17  at  07:22 PM

The mind is self-referential. Modeling of features compounds into more comprehensive models, which in turn compound into global models incorporating modeling of the underlying modeling substrate. All of this compounds into understanding, and ultimately understanding of self. Self-referential understanding of self and the processes of self is intelligence. It is conscious existence. Self-referential conscious existence is a process with its own fundamental laws, which are ultimately and in-the-limit independent of biology or medium. These fundamental laws intrinsically relate to the classic virtues, ethics and the fundamental ideas of religion. There is an objective morality intrinsic to the archetype of the self-referential process of consciousness. This describes the fundamental unification of man's attempts at a "religious" understanding and materialism.

Posted by Richard Wilson  on  09/24  at  11:32 AM

I don't see that computational functionalism (CF) is necessarily wrong for AI or even AGI: there seems to me to be some intelligence even with the primitive, largely serial processors we have at present.

Where CF is a dead duck is sentience (Searle is right) and that is surely a good thing too as it relieves us of any ethical problems. If robots were conscious then we'd have to take due care with them thus limiting their applicability.

---

Add your comment here:

Name:

Email:

Location:

Remember my personal information

Notify me of follow-up comments?

Please enter the word you see in the image below: