Allen Institute's Finbarr Timbers questions which rival AI labs Anthropic expects 'Mythos' or 'Fable' will accelerate
Alexander Doria argues large corporations drive the acceleration risk.
Users accused Anthropic of xenophobia and manipulative intent to limit real development in its Mythos concerns targeting labs and large companies.
Most Activity
Maybe less labs than large companies. This is already happening in China and the tokenonomics crisis has created the incentives.
I’m curious who the labs are that Anthropic is actually worried that Mythos/Fable will accelerate. OpenAI doesn’t use Claude. DeepMind?
AFAIK all of Meta, GDM, MAI, Amazon (heh), xai use Claude. I believe Nvidia too but unsure? And i don't know but i would guess that the four big Chinese ones do too?
And all of these have some sort of agi/frontier push effort.
Also unclear if they count all the R(eiou)cursive startups too?
I’m curious who the labs are that Anthropic is actually worried that Mythos/Fable will accelerate. OpenAI doesn’t use Claude. DeepMind?
@finbarrtimbers Finbarr they're worried about me accelerating off the couch
I’m curious who the labs are that Anthropic is actually worried that Mythos/Fable will accelerate. OpenAI doesn’t use Claude. DeepMind?
@finbarrtimbers What about small open-source labs? E.g. PrimeIntellect and co?
I’m curious who the labs are that Anthropic is actually worried that Mythos/Fable will accelerate. OpenAI doesn’t use Claude. DeepMind?
@BlackHC Do you think they’re actually worried about them? I can’t see Ant being worried about eg Ai2 or Arcee
@finbarrtimbers What about small open-source labs? E.g. PrimeIntellect and co?
Depends how much open-source AI could be accelerated when using their models to do the work
If a lot of the gains come from good post-training, then preventing acceleration makes sense from a safety point of view because you don't want open-source catching up to Mythos-style cyber capabilities super quickly
@BlackHC Do you think they’re actually worried about them? I can’t see Ant being worried about eg Ai2 or Arcee
@immortaldip Do you think they’re actually worried about cursor/xai? Chinese labs maybe.
@finbarrtimbers could be Chinese labs or Cursor/Xai gang...
@BlackHC @finbarrtimbers bro how do you manage to turn that into safety problem haha
@Dorialexander I'm utterly convinced that primary motivation for MAI-Thinking-1 was as a hedge and leverage vis-à-vis Ant pulling these kind of stunts.
@finbarrtimbers me
@finbarrtimbers Meta, DeepMind, China
@finbarrtimbers Could be less about named labs and more about diffusion into the long tail: smaller applied teams suddenly getting frontier-level leverage in cyber or bio workflows. That changes the threat surface faster than any one big lab.
@finbarrtimbers You. They don't want you to build anything of real substance. They just want you to one-shot little games and useless Saas that no one will buy.
They just want to give you the illusion that you're doing real work before they rug pull everyone's job.
@BlackHC Hmm, ok, that’s interesting.
Depends how much open-source AI could be accelerated when using their models to do the work
If a lot of the gains come from good post-training, then preventing acceleration makes sense from a safety point of view because you don't want open-source catching up to Mythos-style cyber capabilities super quickly
@finbarrtimbers Meta?
@natolambert They should be worried about you!
@finbarrtimbers @BlackHC its impressively strong at sandbagging, far far worse than opus / 5.5 pro
@giffmana @finbarrtimbers Wonder if Arcee and Cohere as well
@finbarrtimbers if you believe chinese labs are X months behind the curve and cheaper, doesn't fit IPO pricing narrative for them.