Within a five years it will be possible to create realistic AI virtual actors and sets that generate the entire video sequence including the performance. We can already do simplistic and less realistic versions of this.
There will always be humans willing to act and humans willing to watch them. Maybe there won’t be 10 million dollar wages for a single role, but so what?
they don't want to make art, they want to make content
Apparently, the masses don't care about art.
Sure, you can ask it to draw a beach, but can you ask it to draw the SAME beach with a character interacting with items on that beach?
The current simplistic versions of this have very tightly bound environments for the characters to interact (I'm thinking endless Seinfeld).
That to me means we are talking about the current state of AI.
I don't disagree with setting a legal framework around ownership of AI derived art. However, I don't think this is a pressing issue. AI art, outside of maybe simple picture generation, is bad. AI books are terrible. And I don't see it getting any better.
I'd be shocked if in the next 5 years there's any hit AI art generated. Heck, I'd be shocked if it happens within the next 20 years.
The Claude 2 LLM just released has 100,000 token context and is fairly intelligent.
Take a look at the output of Stable Diffusion XL 0.9 versus Stable Diffusion 1.5. It's a massive improvement.
Metahumans is on an early version and already fairly realistic. Look at the realism of a game like Unrecord.
AI is getting much better very rapidly.
Citation needed. How is IQ being measured? What was it previously?
OP equates number of tokens with intelligence, I do not.
OP then mentions a hyper realistic video game. Cool, but also not exactly evidence that AI could be plopped on top of that to make Jerry Springer.
Throughout the years, the one thing that has remained unchanged is that the power ultimately lies in controlling distribution.