Hacker News new | past | comments | ask | show | jobs | submit login

> No computer will ever demonstrate actual intelligence, which is the ability to bring a new and unforeseen solution to an existing problem

It's easy to disprove absolute claims such as yours. Yes AI can create new things and it's only getting better. You just need 1 example. Read up on AlphaGO. It was able to beat humans by coming up with techniques that no human has ever and possibly would never have come up with because of how illogical they seem to even the masters of Go.

Look into GANS. AI can create new things.




Contemporary AIs don't create new things via creative insight or intuitiveness. A trained NN is only good at the one thing it has been trained at. That isn't true intelligence, no matter how astounding the results.


You're responding to someone mentioning AlphaGo, surely you've heard of AlphaZero. A trained NN can be good at several things.

Creative insight or intuitiveness… What are those exactly? Intuitiveness is so vague of a concept I feel it could accurately describe how some NN behave on the surface, if you didn't know how they are implemented. Intuitiveness in humans is just hidden layers.

Corollary: if you had perfect knowledge of how the human brain work you wouldn't find it so magical and it wouldn't feel so out of reach to replicate its high level emergences.


> A trained NN is only good at the one thing it has been trained at.

I don't know about you, but I only learnt how to read and write thanks to being taught (trained) by other humans.

In contrast, AlphaZero managed to master the games of Chess, Shōgi and Go without ever seeing a single game played, just by being told the rules once, and playing against itself.


The other way of looking at it is that AlphaGoZero was provided with perfect information consisting of every relevant objective fact about the game of Go up front, and thus had perfectly targeted success criteria for the millions of iterations AlphaZero needed to run to get good at it.

I didn't need this to be able to start learning to read, and if the human race learned in this manner, written language wouldn't exist.

And I'm not even sure how one would begin to assemble all relevant information about the space "what words and syntax actually mean" which is a bit more difficult to define than a Go board, or a success function that could describe optimal understanding of books in order to teach computers how to intepret language like AlphaGoZero learned to estimate a mathematical optimization problem in disguise as a board game. Unless your criteria for 'learning to read' is satisfied with parsing text to perform Chinese room tricks in selected domains, or satisfied with textual insights on the level that Dr Seuss appears to have been unfavourably disposed towards green eggs and ham!


I actually taught myself the skill of reading. Real reading, including understanding what the words and sentences mean. While learning countless other things that 5-year-old kids usually learn, partially on my own, partially taught by my parents.

And almost all of that stuff was way more useful than mastering Chess or Go.


The means to do those things were programmed by humans. While the computer is faster at assessing all of the possibilities, it isn't coming up with anything that wasn't made by possible by the human-programmed software.

We now have to have safety belts for Amazon workers to keep them from being run over by robots!


It's unclear to me if you are familiar with GANs, mentioned by grandparent.

A software can hallucinate entirely new photographs that did not exist before of a certain category of objects, like a "dog" or a "burger" or a "celebrity".

Of course it was "made possible" by the human-programmed software, but very very indirectly. In particular, the programmer doesn't really know deeply how it does its thing, and cannot possibly create the same result himself manually.

https://www.youtube.com/watch?v=ZKQp28OqwNQ




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: