@210de87f I've seen this before and it bears a rewatch but you know what strikes me this time?

That they didn't have a problem to solve. They couldn't seem to explain why they had been working on this. That's why they seemed to be grasping at straws as to what to do with it. This on the week that Microsoft is looking to build their own nuclear reactors to power their language models they call "AI", which nobody really wants. Because it is a solution looking for a problem. 

 @7518be2f 👆 this.

In case of OpenAI, Microsoft and Google, the whole shtick is "convince everyone that LLMs — which only gigantic companies can train and maintain — are somehow awesome and inevitable, scare lawmakers with 'superintelligence explosion' into regulating competition away, corner the market".

It's neither awesome nor inevitable.

@210de87f 

 @e10c41d2 @7518be2f Thing is that LLMs are actually pretty good at some things. I've been using Bing Copilot for searching rather than Google for quite a few things, and it works really well when you want to ask a question and then refine it or do follow-ups (and it provides reference links for everything, which is very important). But that's not creative – and LLMs are just baaaad at creative work.