⿻ Andrew Trask

@andyreed @Miles_Brundage it's too good

@Miles_Brundage Burning Min

I think I would really enjoy Burning Man For five minutes Who's building this

@AlbertBuchard care to expand?

@iamtrask Nah..

jobs which are safe from AI: taste, trust, and rare data

@badcryptobitch awww mate! hope you feel better soon!

TFW you were trying to ship for @iamtrask but you ended up in the hospital instead 😅

@badcryptobitch @StoffelMPC

@iamtrask I'll need to read this! Your thoughts on federated architectures aligns with what we've been working on @StoffelMPC and some of the initial products we need to launch to get the ball rolling in terms of protecting various forms of various online in the face of AI

@badcryptobitch internetat50.com/references/Lic… the paper i mentioned (super cool work!)

@badcryptobitch totally agree! (some thoughts)

@l_donciccanplay x.com/iamtrask/statu…

@l_donciccanplay on it

@iamtrask AI is an utility like a knife for example. A shop can't charge customers based on what do they do with the knife but AI companies could charge based on output; text, homework, image, video, code, or research (material discovery), etc.

subject: why intelligence won't be a utility (cuz cryptography)

@zhangbaiwu @openminedorg

@iamtrask @openminedorg Input tokens cost money too, not just output. But I genuinely agree that market won’t like this. Last time o checked utility companies have a much lower multiplier than OpenAI

@sebkrier awww... you're too kind *blushes*

@iamtrask Why! I was half zombified that evening!

I've developed the unique skill of being incredibly sharp in meetings that don't matter, and a half-zombified wreck when surrounded by people who actually do. Bodes well.

Fwiw - JCR Licklider's foundational vision paper for the internet described communication as the alignment of mental models between people. internetat50.com/references/Lic… And pretty much everything that humans do together involves communication... so yeah alignment is pretty much everywhere.

in the end, isn’t everything simply an alignment problem?

@Miles_Brundage Fwiw, best I can tell rank-and-file staff at each lab don't hate staff at other labs. If animosity exists, it would only be limited to a microscopically small group of people (<5?). There's simply to much back-and-forth talent swapping for the former to have developed.

If you think a US-China AI race creates big risks of safety accidents caused by corner-cutting, you should also worry about the leaders of American AI companies hating each other + assuming the worst about each other's intentions

What if instead of sorting AI policy out via threats and vague announcements between the executive branch and companies, we elected people to a branch of government that wrote *public* rules + voted on them? They'd congregate periodically - ooo we we could call it a Congress

@garrytan The internet had a pretty good retention curve, no?

Does a hype cycle have a retention curve that looks like this? Literally never seen a retention cohort graph like this