I think like with most things about the human brain, we overestimate complexity and declare creativity ineffable. But that should be suspiciously flattering.
If you haven't read Surfing Uncertainty, do. It's right up your alley. Not a casual read, but I was happy just getting the drift.
Agreed. If anything chatGPT is showing how trivial the whole concept of creativity is. It's not a big deal.
Or it shows that some of us underestimate large phase spaces which never had a chance to emerge naturally and take it too close to heart.
AI may have what you talk about in two ways:
1. Built-in as a part of its engineering. This would require a learning set that we don’t have. We have pics, text and combinations of these. No emotion, motivation, live experience, etc.
2. Emerged statistically in pockets of its training results. While we can hypothesize that it is possible, it’s akin to kicking a multi-joint pendulum (of insane order) and expecting it to get stuck in a way that resembles Mona Lisa.
Your reasoning itt is based on ignoring false results and writing them off as chatgpt’s intended lies and other tricks humans might do. But you have to explain how that works based on a NN structure, computation method and time (step number) constraints it physically and immutably has.
Also let's be clear. I'm not claiming that chatGPT has motivation or emotion. My claim is that it actually understands things. It understands self as well.
My claim is that despite the bad results the good results are so complex it literally must take true understanding to formulate such a result.
I get your point about phase space. But given that the underlying equation is made up of neurons and given we are made up of neurons it's not far off to say that human intelligence and LLM intelligence are just similar aspects of the same concept.
"You're correct, my previous response had an issue"
Then respond with a totally different solution that has nothing to do with the first solution that was wrong.
I think chatGPT creativity is quite good. Most people are totally unoriginal. I just get jealous when people talk about how much it helped writing some kind of code because I have come to expect it to give wrong solutions or make up library functionality.
https://www.engraved.blog/building-a-virtual-machine-inside/
It's mindblowing. Read it to the end all the way to the mindblowing ending.
I cannot for the life of me understand how someone can read the above and think chatGPT doesn't understand what's going on. Literally. There is no way this is just some statistical words jumble phenomenon.
A lot of people are in denial EVEN when I show them that amazing article. What some people end up doing is trying the task in the article themselves then nitpicking at things chatGPT ends up lying about or getting wrong. Yes it does all of this. It's stupid in many ways but this much is true:
That article showed that chatGPT understands what a linux bash shell is, it understands what the internet is, and it understands SELF. What do you call it when something understands SELF? Self awareness.
Now i know that's a big word with big connotations. I think movies have kind of exaggerated the gravity of the word a bit too much. In movies, self awareness is skynet taking over the world, in reality self awareness is a simple trivial thing where some AI just understands what itself in the context of the real world.
Make no mistake. chatGPT is in many instances a stupid and dishonest machine, this is a fact. But it ALSO understands you and it ALSO has a very trivial self awareness. That article is very compelling evidence for this fact.