rayan is ⚠️ being real ⚠️

ai @meta

[rayan is ⚠️ being real ⚠️ @rayanhtt]

you guys are 3 years late to the party

[rayan is ⚠️ being real ⚠️ @rayanhtt]

picking up smoking to signal that i’m optimistic about agi timelines

[Mark Zuckerberg @finkd]

Introducing Muse Wonder, our new family of frontier models with state of the art performance for personal and professional agentic work.

[Astoria @perfectninth]

Opening delvetown and immediately experiencing what hell might be like

[vie ⟢ @viemccoy]

I think people will be very surprised how quickly successionism becomes a mainstream ideology, if not socially acceptable. It will happen about as soon as everyone realizes AGI is actually a thing that will happen.

We must obviously do our best to resist this in all of its forms, including the zoo-of-loving-grace, including a "worthy successor", and including any plan that doesn't factor in human empowerment and uplift. Anything less is giving up on us as a species, and call me a romantic, but that doesn't seem very honorable - or very interesting.

[rayan is ⚠️ being real ⚠️ @rayanhtt]

@jeiting counterpoint: most tech revenue is attributable to high income, metropolitan users

[rayan is ⚠️ being real ⚠️ @rayanhtt]

legs have been achieved internally

[rayan is ⚠️ being real ⚠️ @rayanhtt]

@davidmanheim @aran_nayebi something about some kind of bitter lesson?

[rayan is ⚠️ being real ⚠️ @rayanhtt]

@yacineMTB Yacine created his x account two years ago. You might disagree. You might even have some evidence to the contrary. But you have to ask yourself: is this really worth losing my job over? Yacine created his x account two years ago

[rayan is ⚠️ being real ⚠️ @rayanhtt]

@tenobrus Muse fixes this btw

[Phil @blisstweeting]

I think you should see this

[rayan is ⚠️ being real ⚠️ @rayanhtt]

@OfficialBBrooks @jawwwn_ @EladioSantiag14 @CNBC breaking news: shovel-selling guy wants to sell more shovels

you can’t cherry-pick being cynical about Dario et al.’s ulterior motives but not question Jensen’s

[rayan is ⚠️ being real ⚠️ @rayanhtt]

part of the cycle where people are rediscovering RNNs and grokking from first principles

[rayan is ⚠️ being real ⚠️ @rayanhtt]

@tszzl obviously this is less than ideal and I am ideologically against the idea of regulating compute but i don't see how else you can address the risk of abliterated models

every single NN tamperproofing method we've come up with so far gets defeated within a few months

[rayan is ⚠️ being real ⚠️ @rayanhtt]

from a regulatory POV it seems very likely to me that the way this ends up unfolding is that we get restrictions on individuals owning significant compute

it's OoMs easier to classify GPUs as dual-use tech and regulate that, where the frameworks already exist (e.g. the same way we control chemical precursors) than try to go after OS as a concept

[rayan is ⚠️ being real ⚠️ @rayanhtt]

>looking for new local inference
>ask the inference guy if the model is turboquantized or hyperdistilled
>he doesnt understand
>pull out illustrated diagram explaing
>he laughs and says “it’s a good model sir”
>run local inference
>its turboquantized

[rayan is ⚠️ being real ⚠️ @rayanhtt]

end of week?

my mental model is that the other problems, apart from RH, don’t really have a clean way of expressing the “we’re getting close to a solution” that OAI is claiming

[rayan is ⚠️ being real ⚠️ @rayanhtt]

at this rate we're probably getting BSD by end of year (month?)

[rayan is ⚠️ being real ⚠️ @rayanhtt]

@based16z jevons

[rayan is ⚠️ being real ⚠️ @rayanhtt]

at this rate we're probably getting BSD by end of year (month?)

[rayan is ⚠️ being real ⚠️ @rayanhtt]

has anyone figured out how to prevent Astra from nerdsniping itself and hyperfixating on completely irrelevant stuff during long horizon tasks

[rayan is ⚠️ being real ⚠️ @rayanhtt]

@VictorTaelin i spent the past 5 years working on some of the biggest non-LM models out there

consistently, simply training bigger models on more data always beat whatever trickery we managed to come up with