I've been having a few surreal conversations about AI with one or two people I know who may be less pessimistic about the tech than I am.
-
I've been having a few surreal conversations about AI with one or two people I know who may be less pessimistic about the tech than I am.
In both cases we are not talking about investor types, these are artists. And they'll say something like "Now I can shoot that period drama I wrote and didn't have the budget for!"
And I want this to be true for them. (In this case I've read the script and it's very good and I would like to watch it.)
1/
-
I've been having a few surreal conversations about AI with one or two people I know who may be less pessimistic about the tech than I am.
In both cases we are not talking about investor types, these are artists. And they'll say something like "Now I can shoot that period drama I wrote and didn't have the budget for!"
And I want this to be true for them. (In this case I've read the script and it's very good and I would like to watch it.)
1/
But, I will say something like "have you tried to get it to do any moment from the script?"
And this is when they say "yeah, it was hard it couldn't get the faces right..." etc. Because that's where this stuff fails.
But then they say "soon it will be better"
It's been three years.
But also, I don't know what we are arguing about? I guess if it did work ... great. But it doesn't.
Maybe they just want me to share the sense of ... hope? It's hard. This has been such a massive mess.
2/2
-
But, I will say something like "have you tried to get it to do any moment from the script?"
And this is when they say "yeah, it was hard it couldn't get the faces right..." etc. Because that's where this stuff fails.
But then they say "soon it will be better"
It's been three years.
But also, I don't know what we are arguing about? I guess if it did work ... great. But it doesn't.
Maybe they just want me to share the sense of ... hope? It's hard. This has been such a massive mess.
2/2
I think they will continue to be disappointed. LLM generated anything is, at best, "good enough". The more specific your need, the less likely an AI is to meet it.
LLMs don't understand anything, they have no underlying model of the world. So they make crazy mistakes that no human would make while creating things that look superficially correct.
-
I think they will continue to be disappointed. LLM generated anything is, at best, "good enough". The more specific your need, the less likely an AI is to meet it.
LLMs don't understand anything, they have no underlying model of the world. So they make crazy mistakes that no human would make while creating things that look superficially correct.
