AI300
  1. Shane Legg / Chief AGI Scientist at DeepMind
    2221.
    @2024dion @deredleritt3r The fact that it still seems 50-50 by 2028 surprises me! I would have thought that evidence would have tipped it one way or another by now.
    LLM score 30 · 6 months ago
  2. Sergey Levine / Physical Intelligence Cofounder
    2222.
    Q-learning with adjoint matching is our latest method for offline RL & offline -> online RL with diffusion/flow models, and (so far?) the best-performing method we've developed for training flow policies with RL.›
    LLM score 95 · 6 months ago
  3. John Carmack
    2223.
    #PaperADay 10 LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristics›
    LLM score 80 · 6 months ago
  4. Boris Cherny / Creator of Claude Code
    2224.
    @ariccio Use it for when the main agent doesn't need to see the work that a skill did, only its output.›
    LLM score 40 · 6 months ago
  5. Michael Elabd / DeepMind Researcher
    2225.
    Honestly, I think memory is the biggest blocker to continual learning right now.›
    LLM score 85 · 6 months ago
  6. Andrew Lampinen / Research Scientist at DeepMind
    2226.
    Is it better to go to academia or industry for research in AI or cognitive science? It's the most frequent question I get asked by PhD students, and I've finally written up some of my thoughts on the answers, as an epilogue to my more research-focused posts on cog sci and AI: https://t.co/qoIqBKGqX6
    LLM score 75 · 6 months ago
  7. Yoshua Bengio
    2227.
    Il est essentiel de développer à la fois une réglementation efficace, des garde-fous sociaux et des solutions techniques comme celles que nous développons à @LawZero_ pour éviter les risques de l’IA avancée tout en développant ses applications positives.›
    LLM score 40 · 6 months ago
  8. Shane Legg / Chief AGI Scientist at DeepMind
    2228.
    @kevinkshah @korbencopy I'd give it a 50% probability of happening within 5 years of Minimal AGI.›
    LLM score 70 · 6 months ago
  9. Shane Legg / Chief AGI Scientist at DeepMind
    2229.
    @GDalegend @korbencopy We get to 2028 and have something quite close to Minimal AGI, but not quite.
    LLM score 20 · 6 months ago
  10. Shane Legg / Chief AGI Scientist at DeepMind
    2230.
    @tablefourthree @BogdanIonutCir2 @korbencopy It's actually surprising to me that my probability hasn't shifted yet, given that 2028 is so close.›
    LLM score 20 · 6 months ago
  11. Shane Legg / Chief AGI Scientist at DeepMind
    2231.
    @korbencopy 50% chance of Minimal AGI by 2028.›
    LLM score 70 · 6 months ago
  12. John Carmack
    2232.
    #PaperADay 9 floq: Training Critics via Flow-Matching for Scaling Compute in Value-Based RL›
    LLM score 90 · 6 months ago
  13. Kevin Lu / Thinking Machines Researcher
    2233.
    encouraging progress on continued test-time adaptation beyond model deployment!›
    LLM score 65 · 6 months ago
  14. Karan Dalal / Stanford SAIL Researcher
    2234.
    Test-time training is inevitable.›
    LLM score 70 · 6 months ago
  15. Ronak Malde / DeepMind Researcher
    2235.
    At Deepmind ever since we maxed out SWE-bench , I’ve felt that there hasn’t been a difficult enough benchmark that is aligned with how people want to use AI.›
    LLM score 75 · 6 months ago
  16. Sergey Levine / Physical Intelligence Cofounder
    2236.
    Benjie posted a reply to the Physical Intelligence "Olympics" attempt: https://t.co/sPkc5DK4ls›
    LLM score 70 · 6 months ago
  17. Siyan Zhao / UCLA PhD
    2237.
    Introducing 💡On-Policy Self-Distillation💡, a simple method that enables LLM to teach itself with dense per-token feedback on its own on-policy generations—achieving 4-8x more token efficiency vs.›
    LLM score 92 · 6 months ago
  18. Noam Brown / OpenAI Research Scientist
    2238.
    @_sholtodouglas @jekbradbury @GoogleDeepMind @andy_l_jones @OpenAI @kevin_wang3290 A word on comp: I know folks who become quants to make money but 5 years later ask what they're doing with their life.›
    LLM score 20 · 6 months ago
  19. Noam Brown / OpenAI Research Scientist
    2239.
    @_sholtodouglas @jekbradbury @GoogleDeepMind @andy_l_jones Labs like @OpenAI also hire researchers straight out of undergrad, like @kevin_wang3290, though the bar is high.›
    LLM score 60 · 6 months ago
  20. Noam Brown / OpenAI Research Scientist
    2240.
    @_sholtodouglas @jekbradbury @GoogleDeepMind @andy_l_jones was a semi-retired quant when he wrote a paper comparing the impact of scaling pretraining vs scaling test-time compute (before TTC scaling was cool).›
    LLM score 70 · 6 months ago
  21. Demis Hassabis / CEO of DeepMind
    2241.
    We’re excited to be working with @JNJInnovation to accelerate the path to new medicines.›
    LLM score 60 · 6 months ago
  22. John Carmack
    2242.
    #PaperADay 8 Beyond Gradient Averaging in Parallel Optimization: Improved Robustness through Gradient Agreement Filtering›
    LLM score 85 · 6 months ago
  23. Igor Babuschkin / Cofounder of xAI
    2243.
    @MillionInt The engineer presented a single grain of etched silicon.›
    LLM score 65 · 6 months ago
  24. Lewis Tunstall / Hugging Face Researcher
    2244.
    This is a big deal because it shows that (a) Claude can solve hard math problems when given an appropriate scaffold with tools (Lean) and (b) the framework generalises beyond competitive mathematics to enable autoformalization at scale›
    LLM score 85 · 6 months ago
  25. John Carmack
    2245.
    #PaperADay 7 Cautious Weight Decay https://t.co/D0u9KL5zsS›
    LLM score 95 · 6 months ago
  26. Keller Jordan / OpenAI Researcher
    2246.
    @supreme__sharma There's no good answer.›
    LLM score 20 · 6 months ago
  27. Keller Jordan / OpenAI Researcher
    2247.
    @eliebakouch In terms of hypothetical disagreement-winrates between these different cultures, I’d estimate that my (heavily muted) FYP has +500 ELO vs.›
    LLM score 30 · 6 months ago
  28. Clive Chan / OpenAI Hardware
    2248.
    nothing is free in computer architecture.›
    LLM score 85 · 6 months ago
  29. Clive Chan / OpenAI Hardware
    2249.
    Dojo1 was ultra ambitious and really pushed a lot of technologies (packaging, power delivery, system design, clock distribution, even number formats), but execution and focus could've been better.›
    LLM score 85 · 6 months ago
  30. Keller Jordan / OpenAI Researcher
    2250.
    The internet has become real life only in the sense that *public* cultural systems have been largely replaced by it›
    LLM score 20 · 6 months ago
More