Winston SmithApr 3
TechAltar

Anthropic's massive oopsie

Nebula: https://nebula.tv/videos/tfc-anthropics-massive-oopsie
YouTube: https://youtu.be/eYunGcpRYks

Apr 3 at 1:20pmWeb
Show threadWinston SmithApr 3
@Techaltar so, can be call Claude open source? 😂
Show threadKeith BöhlerApr 3

@dm29 @Techaltar

Only if you run it by another AI. This will satisfy the idea that the work is now transformative. The AI companies themselves have used this to separate their outputs from training data. Hence the python translation.

This is a very meta layered April Fool joke.

Show threadPinkApr 3
@Techaltar you said you don’t think there will be a single moment where one model suddenly becomes much better than the others. I’d like to challenge that by saying: whoever first manages to build an LLM that essentially has an unlimited context size because it can re-train itself during inference, will suddenly be leaps ahead of the competition. Thoughts? (Maybe for next week’s Q&A?)
Show threadTechAltarApr 3
@can I don't know enough about the tech to answer definitively 🤔
Show threadPinkApr 3
@Techaltar that's fair, I respect that
Show threadolbotta4d ago
@Techaltar question for the next episode: what do you think about the recent comeback of pebble watches? What do you think about the hardware and software?