Another Article About Singularity

http://www.technologyreview.com/view/...

---

I believe that the brain is computable, but that it would require emergence out of an extremely complex system.

There's is a poor analogy in this article--Nicolelis is saying that the stock market can't be predicted because it's not computable, which is wrong. The stock market IS computable, because all those transactions are tracked by computers and the stock market exists as a digital entity--that no one has been able to model it yet to allow for accurate predictions does not mean that it's not computable or that it will never be predictable, it just means that it is a digital entity that has outputs we do not fully understand. But the entity already exists.

Given enough time, I suspect that the stock market will generate enough of a history of big data that fast-computing and improved algorithms will be able to model predictions with increasing accuracy based on trending + financial data (and barring external events like scandals and natural disasters which don't have to do with the stock market anymore but affect it).

Nobody could fathom predicting the weather before. Now, we can. It's not perfect, but the accuracy is increasing.

So:

1) He's wrong about the stock market.
2) The stock market is a totally different problem space from intelligence.

What he is really saying is that intelligence requires a size-N network of interactions and that no computer will every be powerful enough to model such a thing.

That is weak. You have to show that a problem is inherently unsolvable, not assume that size-N is unsolvable because nothing today is powerful enough to run the N-sized network. Computers are STILL growing in speed more and more; every time a cap with current tech is reached a new refinement or breakthrough gets beyond that limit. Hardware powerful enough to simulate a human mind may exist by the end of the century.

So I disagree! Also, Stephen Hawking disagrees with you too, Mr. Nicolelis.

---

TL;DR:

I still don't think the Singularity will happen though, for reasons I've mentioned before.

(P is not NP; intelligence is a problem of such complexity that each step up involves a massive increase in problem difficulty; iterative acceleration of intelligence development will not happen, it will level off).

But having a strong general artificial intelligence? I think that's definitely doable. And it will change a great deal about the world.

---