Not as funny as reality though, where AI is suddenly good at art and writing, but still bad at flipping burgers, or simply navigating floor of a house.
> Moravec's paradox is the observation by artificial intelligence and robotics researchers that, contrary to traditional assumptions, reasoning requires very little computation, but sensorimotor and perception skills require enormous computational resources.
Moravec's paradox is just an observation made on 80's state of AI. It doesn't hold true - high-level reasoning is absolutely not cheaper than sensorimotor stuff, and most importantly there might be no difference between them.
> Moravec wrote in 1988, "it is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility".
That is, few compute resources are needed for a program that can beat human checkers players, as compared to what’s required to solve perception problems.
I wonder if you have a different thought of what “reasoning” was being discussed, as you switched the term to “high level reasoning”?
I forget who recently observed this: AI is doing well at art and fiction writing because those don't have to be "correct" but it's less useful for writing code or non-fiction.
I feel like OP is almost claiming the opposite. Art isn't clearly defined - there's no "correct art". Therefore whether you prefer art generated by an AI to art generated by a human artist is, within some margin in a blind test, random.
If I have a human and an AI go illustrate a manual for assembling my new sofa, or produce the advertising materials for the grand opening of a local restaurant, it's much easier for the AI to clearly produce an incorrect illustration in a way a human generally wouldn't.
Most of the "art" created by AI is on the level of assembling a sofa. And sure, lots of people have poor taste and criteria for understanding art.
If nothing is being communicated, there is no art, in my opinion. Until AI is conscious, that will continue to be the case.
Sure, but if you see a piece of art with no context for its creation, you have no way to know what the artist intended to communicate or if they intended to communicate anything at all, or if there was even a human artist involved. I'm aware that Mondrian was a real human who made art, but if you told me Mondrian was an AI, would that make my enjoyment of their work less valid?
In short, I mostly agree that good art is communicative, but also that some of that is because of the context where someone shows you something and tells you it's communicative and maybe what it's intended to communicate. There's plenty of famous works of art that don't look that different than programmatically generated NFT works, let alone things Stable Diffusion does.
People are really good at generating meaning and patterns from actually literally nothing. See conspiracy theories and the gambler's fallacy. If you know it came from an AI that intends nothing, it's easy to see the seams that tell you it's all pieces of meaningless nothing. If you're told that it's from some artist with a long history and passion for expressing some message, it's easy to read communication into anything.
Not as funny as reality though, where AI is suddenly good at art and writing, but still bad at flipping burgers, or simply navigating floor of a house.