Comment by NitpickLawyer
Comment by NitpickLawyer 2 days ago
> So what we have is a hype on top of beliefs in progress as continious phenomena. But progress itself has slowed greately.
I think we can split the two.
a) I don't think there's anyone seriously doubting there's a lot of hype going around. But that's to be expected. Trust not the trust me bros, and so on. I think we agree here.
b) Progress hasn't slowed, and at the same time progress (in research) can stop tomorrow and we'd still have work to do for 20years, simply using what's on the table today. LLMs have solved "intent comprehension". RL seems to have solved "verifiable problems". The two can and will be used together to tackle "open ended questions". And together we'll see advancements in every aspect of our lives, on every repetitive mundane and boring task we have ever done.
> on every repetitive mundane and boring task we have ever done.
A lot of them could be solved with pre-ai things. Many were self-inflicted by people badly designing and approving existing processes. I really don't see how AI is going to get us out of this one. I've been paid money to automate a few existing self-inflicted repetitive, mundane and boring tasks and the companies that created them are not even interested in talking about solving that. Some work the opposite way - in the US the value keeps being extracted from the tax filing system for example, even though we know how to remove the issue itself.
It's weird to discuss how the AI will automate everything when we're actively fighting simplification right now.