Loading activity…
@_arohan_
co-founder, board @coreauto - and continuing to aspire to understand deep learning.
AI
A user argues that the shale revolution's contribution to AI has been underappreciated.
AI
The speaker says they may have a unique perspective from scaling many generations of training systems.
AI
A user jokes that one frontier lab is preparing a better model while others are reportedly releasing better successors.
AI
A user argues that evaluations are central to selling data to frontier labs and building self-improving models.
AI
SemiAnalysis says an upstream vLLM optimization made the 2.8-trillion-parameter model 1.68 times faster on the same hardware.
AI
One commenter says it's hard to separate distillation's contribution from gains made through internal reinforcement learning.
AI
A user compares that potential shift to writing and the printing press, which spread intellectual output widely.
AI
A user says Baguettotron's feature connections differ sharply from Gemma-270m's, perhaps more because of synthetic pretraining than deep architecture.
AI
A shared article excerpt says OpenAI leaders described possible tactics including offers of money and attempts to seduce employees.
AI
A post links to a photo without offering any explanation beyond 'hmm.'
AI
The presenter says the San Francisco talk will cover hardware codesign and synchronous versus asynchronous training.
AI
A user says Argon often doesn't reason at all or leaks all of its reasoning as plain text.
AI
Bespoke says Nimble-v4-max scored 68.4 on the Decision Index, 10.50 points above Jev 1.13.
AI
Google says Argon is available through its paid API tier and to Pro and Ultra subscribers in Gemini and Search.
YouTubeAI
A user calls research “the golden goose” while quoting another post, without elaborating.
AI
Axios reports the move followed officials' claim that Anthropic used the State Department's immigrant visa application system 'fraudulently.'
AI
A poster jokes that Matej's draft claims a Coreauto member-of-technical-staff role focused on reinforcement-learning performance.
AI
A user compares labs’ release language to calling something “the best model in the north western jersey nowadays.”
AI
A poster wonders whether all ambitious neo labs should find ways to work together.
