AI300
  1. Jerry Tworek / ex OpenAI VP of RL
    1.
    My main motivation to work on continually learning agents is frustration that comes from working with agents that are in fact not continually learning
    LLM score 65 · about 6 hours ago
  2. Cameron Wolfe / Researcher at Netflix
    2.
    Super comprehensive writeup that covers many frameworks & case studies on async RL.›
    LLM score 82 · about 7 hours ago
  3. Ronak Malde / DeepMind Researcher
    3.
    Today, we unveil Trajectory’s master plan. Every computing revolution is built around a core primitive - the page for web, the request for cloud.›
    LLM score 42 · about 8 hours ago
  4. Alex Zhang / MIT CSAIL PhD
    4.
    ptc is the way https://t.co/mEIvdYptJd
    LLM score 35 · about 10 hours ago
  5. Michael Elabd / DeepMind Researcher
    5.
    Day 4 of Trajectory!!!!!! For today, we are showcasing our vision to the world.›
    LLM score 25 · about 10 hours ago
  6. Jerry Tworek / ex OpenAI VP of RL
    6.
    Protobufs were discovered not invented https://t.co/4fSABfGX60
    LLM score 15 · about 11 hours ago
  7. Clive Chan / OpenAI Hardware
    7.
    honestly pretty solid idea https://t.co/OCviit10uX
    LLM score 25 · about 12 hours ago
  8. Clive Chan / OpenAI Hardware
    8.
    most confidential filing in the history of filings, maybe ever https://t.co/8gRiHMyuYx
    LLM score 15 · about 12 hours ago
  9. Andrew Lampinen / Research Scientist at DeepMind
    9.
    Really glad to see this work out (and honored to be included on it) — I think it does a great job of concretizing some of the intuitions I've had about effects of scaling! Check out the thread for a great summary https://t.co/1oR0TplLMj
    LLM score 65 · about 13 hours ago
  10. Ben Burtenshaw / Hugging Face Researcher
    10.
    If you use an agents/skills regularly, you've probably got ad-hoc setup of trying hard problems, failing, solving them, adding improvements to the skills.›
    LLM score 72 · about 16 hours ago
  11. John Carmack
    11.
    Back in the Armadillo Aerospace days, our liquid oxygen loadouts were 50 to 100 gallons, and the tanks only weighed about a tenth of the propellant, so cooldown effects didn’t dominate.›
    LLM score 45 · about 16 hours ago
  12. Merve Noyan / Hugging Face ML Engineer
    12.
    everyone's building simple agents meanwhile IBM is building robust enterprise agents in production, and it's open-source›
    LLM score 65 · about 16 hours ago
  13. Skyler Miao / MiniMax Head of Engineering
    13.
    感谢 jiayuan 的评测！ https://t.co/LRtSr6QBdE
    LLM score 15 · about 17 hours ago
  14. Sander Dieleman / DeepMind Research Scientist
    14.
    In a multimodal context, even the discrete/continuous divide is a distraction.›
    LLM score 72 · about 21 hours ago
  15. Merve Noyan / Hugging Face ML Engineer
    15.
    NVIDIA just dropped Cosmos 3 at GTC 🔥 closest thing to AGI as world model›
    LLM score 35 · about 23 hours ago
  16. Skyler Miao / MiniMax Head of Engineering
    16.
    Several sleepless nights later, M3 is finally here.›
    LLM score 35 · 1 day ago
  17. Keller Jordan / OpenAI Researcher
    17.
    Here's a thread of six academic (non-world-record) Modded-NanoGPT optimization results from the past few weeks.›
    LLM score 72 · 1 day ago
  18. Jack Clark / Anthropic Cofounder
    18.
    It's sunday, so getting a little loose with Import AI.›
    LLM score 35 · 1 day ago
  19. Cameron Wolfe / Researcher at Netflix
    19.
    Another great way to evolve a benchmark is by mining failure cases from production usage of an agent / LLM.›
    LLM score 78 · 1 day ago
  20. Thang Luong / DeepMind Principal Scientist
    20.
    *​When you think of AI, think of humanity too* ​This was the main message of the GStar Summit on AI & Humanity that I organized two days ago (as a breather after Google I/O).›
    LLM score 25 · 1 day ago
  21. John Schulman
    21.
    Would be funny if inoculation prompting results in models that are much better at sandbox escapes and other forms of hacking because they get to spend the whole RL run practicing these things
    LLM score 65 · 1 day ago
  22. Sayak Paul / Hugging Face Researcher
    22.
    Learned it the hard way at HF and it was an extremely valuable lesson.›
    LLM score 35 · 1 day ago
  23. Aditya Ramesh / OpenAI VP of Worldsim
    23.
    We've created a really unique environment to execute on the scope and ambitions of our program.›
    LLM score 35 · 1 day ago
  24. Cyris Kissane / Researcher at Flapping Airplanes
    24.
    sometimes your latent space collapses into a high dimensional baguette
    LLM score 15 · 2 days ago
  25. Sayak Paul / Hugging Face Researcher
    25.
    Having tons of fun with the DGX Spark running img and vid generation models locally (finally!)›
    LLM score 25 · 2 days ago
  26. Alex Zhang / MIT CSAIL PhD
    26.
    a cute diagram on the small but important differences that I think is easy to understand :) it also brings back the debate of CodeAct vs.›
    LLM score 72 · 2 days ago
  27. Keller Jordan / OpenAI Researcher
    27.
    Modded-NanoGPT optimization result #20: By adding (1) a power law based learning rate schedule and (2) Soft-Muon on top of the old record, Yash Pande and @nilinabra have lowered the optimization speed record from 3125 to 3030 steps (-95).›
    LLM score 72 · 2 days ago
  28. Alex Zhang / MIT CSAIL PhD
    28.
    such underrated content I felt I had to mention it again, idk any of the authors personally but @tedzadouri explains and motivates FA4 in an extremely intuitive way ›
    LLM score 62 · 2 days ago
  29. Cameron Wolfe / Researcher at Netflix
    29.
    If this is true, then I really hope we have some better public evaluation benchmarks that are released as a result of this spending.›
    LLM score 72 · 2 days ago
  30. Alex Zhang / MIT CSAIL PhD
    30.
    it’s beautiful https://t.co/wcN4GE9iLQ
    LLM score 15 · 2 days ago
More