Hacker News Reader

Comment by RugnirViking | Hacker News Reader

> People don't realize there's an aspect to creativity that dwells within the human brain and the morass of personal visceral human experience that can't be reconstructed

Your point here seems to be based more on faith than anything concrete. It is a sibling with the idea humans have "immortal souls" and the like - something we would like to be true more than anything that actually is.

This being said, I do think AI systems are very bad at creativity. But I don't think that it's impossible that they could be, nor do I have any idea why, specifically, they are bad at creativity. At this point in interpretability research, anyone telling you they do know is lying.

View on HN

>Your point here seems to be based more on faith than anything concrete. It is a sibling with the idea humans have "immortal souls" and the like - something we would like to be true more than anything that actually is.

No, I'm not invoking the concept of an immortal soul, and it's incredibly frustrating that this is the go-to assumption anyone makes when criticism of AI's creative capabilities.

It's entirely possible to be a pure physicalist while also believing AIs aren't capable of creativity because they aren't capable of being human, and don't have the necessary human experience to inform the processes they simulate, because creativity is fundamentally a means of human communication. That isn't actually an argument from the supernatural, and unless you believe it's possible to derive a brain simulation of Hayao Miyazaki complete with his lived experiences and memories simply by studying the specific technical aspects of his animation style in aggregate, then I think the why of AI's limitations in this regard should be obvious.

The problem is that they used to say that about chess. In the 1980s it was thought that grandmaster-level chess wasn't just algorithically looking at the board and computing scores for moves as chess programs did but something that required the intuitive powers of the human mind. After all, it's clear that Kasparov wasn't computing the scores of millions of possible moves across future board states in his head but simply "saw" the right one to take by a process he couldn't explain. And yet he was eventually beaten by a machine - not even one using a clever algorithm - just a machine that did pretty much what any stupid chess program did but faster. It turned out that human "intuition" wasn't needed.

Reply on news.ycombinator.com