Jan Leike

@janleike

Senior Research Scientist at DeepMind, working on agent alignment. Obviously all opinions are my own.

Vrijeme pridruživanja: ožujak 2016.

Tweetovi

Blokirali ste korisnika/cu @janleike

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @janleike

  1. Prikvačeni tweet
    20. stu 2018.

    The agent alignment problem may be one of the biggest obstacles for using ML to improve people’s lives. Today I’m very excited to share a research direction for how we’ll aim to solve alignment at . Blog post: Paper:

    Poništi
  2. 16. pro 2019.

    Lucas Perry from FLI interviewed me for their podcast! If you want to find out why I don't do theory research anymore, what's going on in safety at DeepMind, and how we're planning to solve the alignment problem, check it out:

    Poništi
  3. 15. pro 2019.

    Rob has a whole YouTube channel dedicated to explaining AI safety ideas for everyone to understand. Definitely worth checking out if you're interested in this stuff!

    Prikaži ovu nit
    Poništi
  4. 15. pro 2019.

    Very accessible explanation of the motivation behind reward modeling by :

    Prikaži ovu nit
    Poništi
  5. 13. pro 2019.

    Below you can see our algorithm's generated trajectories in 's Car Racing task. They don't need to be 100% realistic, just good enough for the human to label them correctly (e.g. driving into the grass is bad).

    Prikaži ovu nit
    Poništi
  6. 13. pro 2019.

    How do you train an RL agent in the presence of unknown, unsafe states without visiting them even once? New algorithm by our intern synthesizes trajectories with a generative model and ask a human to label them for safety.

    Prikaži ovu nit
    Poništi
  7. proslijedio/la je Tweet
    10. pro 2019.

    Want to ensure AI is beneficial for society? Come talk to like-minded people at the Human-Aligned AI Social at , Thursday 7-10 pm, room West 205-207.

    Poništi
  8. proslijedio/la je Tweet
    20. kol 2019.

    I love all of the programmes, but I particularly like ep 4 - one of my favourite bits of the whole series is when explains the human preferences work we did with

    Prikaži ovu nit
    Poništi
  9. 14. kol 2019.

    How do you design agents that don’t have an incentive to tamper with their reward signal? et al. derive design principles for RL algorithms. Easy fix if you’re doing model-based RL!

    Poništi
  10. proslijedio/la je Tweet
    19. lip 2019.

    Excited to finally share this AI Reading List, compiling key resources on artificial intelligence & its long-term implications. The list is divided into "80/20" resources and "deep dive" resources to help with suggested prioritization.

    Prikaži ovu nit
    Poništi
  11. 9. lip 2019.
    Prikaži ovu nit
    Poništi
  12. 9. lip 2019.

    Very excited to deliver the tutorial on tomorrow together with ! Be prepared for fairness, human-in-the-loop RL, and a general overview of the field. And lots of memes!

    Prikaži ovu nit
    Poništi
  13. proslijedio/la je Tweet
    3. lip 2019.

    I had a fantastic conversation with scientist Pushmeet Kohli about how they're developing new ways to keep AI systems robust & reliable, why it’s a core issue in AI design that everyone has to attend to, and how to succeed as an AI researcher:

    Poništi
  14. 28. ožu 2019.

    How do we uncover failures in ML models that occur too rarely during testing? How do we prove their absence? Very excited about the work by ’s Robust & Verified AI team that sheds light on these questions! Check out their blog post:

    Poništi
  15. proslijedio/la je Tweet
    19. velj 2019.

    New version on arXiv of our ICLR paper with , , , Edward Hughes, and . We jointly learn language-conditional policies and reward models. Updated results/explanations + discussion of relation with other IRL methods.

    Prikaži ovu nit
    Poništi
  16. 29. sij 2019.

    Recent evidence that there can be unexpected unaligned agents in your data center:

    Tweet je nedostupan.
    Poništi
  17. proslijedio/la je Tweet
    22. sij 2019.

    Join us and this Thursday at 6:00pm GMT for an exciting demonstration, hosted by and ! Livestream on YouTube: Read more about as an environment for AI research:

    Poništi
  18. 18. pro 2018.

    Multiparty computation is awesome because it lets multiple parties train a model without seeing the weights. But there are fundamental limits to making it scalable: >24x overhead! Our new paper addresses this problem. w/ et al.

    Poništi
  19. proslijedio/la je Tweet
    6. pro 2018.

    Good morning ! Stop by our recruitment stand from now until 9:50am to meet with our Safety team. Read more about their work: Later today, our Science team will have a meet & greet at 3pm - come chat about protein folding!

    Poništi
  20. 16. stu 2018.

    Finally, more evidence that the reward model needs to be trained with humans in the loop; otherwise the agent learns to exploit the reward model, for example by pretending to shoot a spider in Hero. 3/3

    Prikaži ovu nit
    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·