AI300
  1. Ethan Shen / Ai2 Researcher
    1531.
    Finally, we conduct an analysis of variance across SWE-Bench runs.›
    LLM score 95 · 4 months ago
  2. Ethan Shen / Ai2 Researcher
    1532.
    We also experiment with mixing data from rollouts and discover that training on SERA’s first and second rollouts complement each other to increase performance in data constrained regimes.›
    LLM score 85 · 4 months ago
  3. Ethan Shen / Ai2 Researcher
    1533.
    Even more interesting is the fact that the best data comes from truncated trajectories that just barely exceed the context limit.›
    LLM score 85 · 4 months ago
  4. Niklas Muennighoff / AI Researcher at Stanford
    1534.
    Community-built open benchmarks work really well, e.g., Terminal-Bench, HLE, MMTEB.›
    LLM score 80 · 4 months ago
  5. John Carmack
    1535.
    #PaperADay 11 Discovering state-of-the-art reinforcement learning algorithms›
    LLM score 80 · 4 months ago
  6. Noam Brown / OpenAI Research Scientist
    1536.
    Had to cut this one for space: 2019: AI can't create art—creativity is uniquely human
    LLM score 20 · 4 months ago
  7. Andrej Karpathy / AI researcher
    1537.
    @0xabi96 It feels like I’m cheating.›
    LLM score 70 · 4 months ago
  8. Andrej Karpathy / AI researcher
    1538.
    @ChiragLathiya The nearest neighbor really is some kind of a junior engineer.›
    LLM score 80 · 4 months ago
  9. Andrej Karpathy / AI researcher
    1539.
    @jeremytwei Love the word "comprehension debt", haven't encountered it so far, it's very accurate.›
    LLM score 30 · 4 months ago
  10. Andrej Karpathy / AI researcher
    1540.
    @airesearch12 💯 @ Spec-driven development It's the limit of imperative -> declarative transition, basically being declarative entirely.›
    LLM score 60 · 4 months ago
  11. Andrej Karpathy / AI researcher
    1541.
    A few random notes from claude coding quite a bit last few weeks.›
    LLM score 20 · 4 months ago
  12. Noam Brown / OpenAI Research Scientist
    1542.
    1987: AI can't win at chess—planning is uniquely human›
    LLM score 70 · 4 months ago
  13. Yoshua Bengio
    1543.
    Had a great discussion on AI's major societal impacts and emerging risks on the @RestIsPolitics podcast this past November.›
    LLM score 20 · 4 months ago
  14. Jonathan Ross / TPU Creator
    1544.
    Success in the Information Age was about being able to answer questions.›
    LLM score 20 · 4 months ago
  15. Ben Burtenshaw / Hugging Face Researcher
    1545.
    this is a blog post on claude + llama.cpp https://t.co/yej6WsNnQA
    LLM score 20 · 4 months ago
  16. Yoshua Bengio
    1546.
    Merci à @HugoDecrypte pour l'invitation à faire un tour d'horizon des risques de l'IA ainsi que des solutions nécessaires pour se diriger vers une meilleure trajectoire.›
    LLM score 20 · 4 months ago
  17. Ben Burtenshaw / Hugging Face Researcher
    1547.
    I know it's massively off trend, but I love setting up a nice quiver of models in @cursor_ai, reviewing generated tokens in the legit ui, switching models for certain parts, and just generally getting shit done.›
    LLM score 30 · 4 months ago
  18. Ronak Malde / DeepMind Researcher
    1548.
    I don't think people have realized how crazy the results are from this new TTT + RL paper from Stanford/Nvidia.›
    LLM score 85 · 4 months ago
  19. Clive Chan / OpenAI Hardware
    1549.
    latest response hit 9h25m + 2.3 million output tokens and did everything i wanted it to over the course of TWELVE context compactions https://t.co/F9ByO19kc0
    LLM score 85 · 4 months ago
  20. Cameron Wolfe / Researcher at Netflix
    1550.
    Continual learning is being positioned as a prerequisite for AGI (i.e., general systems must be adaptable).›
    LLM score 70 · 4 months ago
  21. Geoffrey Hinton
    1551.
    I just watched a really great conversation about the future of AI.›
    LLM score 30 · 4 months ago
  22. Ben Burtenshaw / Hugging Face Researcher
    1552.
    The only thing I don't get is why it has to be a Mac mini.›
    LLM score 20 · 4 months ago
  23. Chelsea Finn / Physical Intelligence Cofounder
    1553.
    Video models serve as a good pretrained backbone for robot policies.›
    LLM score 70 · 4 months ago
  24. Sander Dieleman / DeepMind Research Scientist
    1554.
    As I said in my blog post about latent representations last year: "representation learning and reconstruction are two separate tasks, and while it is convenient to do both at once, this might not be optimal."›
    LLM score 85 · 4 months ago
  25. Shane Legg / Chief AGI Scientist at DeepMind
    1555.
    @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 · 4 months ago
  26. Sergey Levine / Physical Intelligence Cofounder
    1556.
    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 · 4 months ago
  27. John Carmack
    1557.
    #PaperADay 10 LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristics›
    LLM score 80 · 4 months ago
  28. Boris Cherny / Creator of Claude Code
    1558.
    @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 · 4 months ago
  29. Michael Elabd / DeepMind Researcher
    1559.
    Honestly, I think memory is the biggest blocker to continual learning right now.›
    LLM score 85 · 4 months ago
  30. Andrew Lampinen / Research Scientist at DeepMind
    1560.
    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 · 4 months ago
More