Cal Newport

underestimated the degree to which when we as humans are interacting with a chatbot that we're incredibly gracious, we're able to adjust and kind of get the gist of what it means and filter out the part of the chatbot response that's not really relevant or ask the follow-up question.

And when they tried to just use LLM responses to automate

actions within programs, it's just not accurate enough.

So they wanted that to be the case, that you could just be talking to a Riverside bot and you never would have to press a button ever again in Riverside.

It's just not accurate enough.

LLMs, it's fine for human conversation.

It's just not accurate enough in this general case.

Yeah, I was going to say, how much did Meta spend on the Metaverse?

Customizing floating dinosaur avatars.

That's the second $50 billion, right?

If they had gotten the second half of the investment, they would have got to the legs.

They're just not there yet.

It's wacky.

I can't believe we're letting people get away with having a psychologist talking through the chatbot in your system card.

It's nuts.

It's all gone so marketing.

Yeah, I mean, FreeBSD Kernel is full of bugs.

All these things are full of bugs.

Because they're open source.

I had to have this conversation with someone recently where they were like, Mythos, can you believe of all the places it found a bug,