Sebastian Risi

@risi1979

AI researcher. Neuroevolution, Human Computation. Associate Professor, ITU Copenhagen. Consultant Uber AI Labs. Co-founder of

Vrijeme pridruživanja: travanj 2009.

Tweetovi

Blokirali ste korisnika/cu @risi1979

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @risi1979

  1. proslijedio/la je Tweet
    prije 53 minute

    Here's our team, in case you were wondering. We're all standing up straight, or at least perpendicular to some high-dimensional hyperplane that is not depicted because of the inherent limitations of projecting onto a 2D plane.

    Prikaži ovu nit
    Poništi
  2. proslijedio/la je Tweet
    prije 1 sat

    We have also hired a team of top-notch talent, including . We are incredibly grateful that these researchers, who have all done remarkable research in their own right, believe in our vision.

    Prikaži ovu nit
    Poništi
  3. proslijedio/la je Tweet
    prije 2 sata

    We're thrilled to announce we've closed 's seed round, raising $1.7M from a fantastic group of investors supporting our mission to bring the best possible AI-driven tools to game developers around the world:

    Poništi
  4. prije 1 sat

    Very happy our startup was able to close this seed round! Looking forward to the cool things this allows us to build! 😎👾🎮

    Poništi
  5. 20. sij

    Hopefully this will motivate more people to do research in artificial life 🦠, instead of following the same old gradients🐾. On the other hand, it's also nice working in a field that's not as crowded yet 🙂

    Poništi
  6. proslijedio/la je Tweet
    20. sij

    for example, the last issue had tweets from discussing accountability, on deep learning theory, on a tool for understanding bias, on using ML with doodles instead of photos, @FidlerSan on a search engine for training data, and more!

    Prikaži ovu nit
    Poništi
  7. proslijedio/la je Tweet

    You know that open offices are working well when a yellow coffin seems like a pleasant alternative

    Poništi
  8. 16. sij

    What conference organizers say "We are extending the paper deadline because of numerous requests ..." What it actually means "We do not have enough submissions and desperately need more..." 😆

    Poništi
  9. proslijedio/la je Tweet
    13. sij

    I love it when scientists embrace both research & education, & try to make their work understandable to general public.

    Poništi
  10. 13. sij

    The tool combines counterfactual examples and clustering of the networks' activation patterns to empower non-experts to detect bias. For most of us, this is the first foray into fair AI methods, so any pointers to other relevant work or suggestions are greatly appreciated.

    Prikaži ovu nit
    Poništi
  11. 13. sij

    Traditionally, non-experts have little control in uncovering potential social bias in the algorithms that may impact their lives. We present a preliminary design for an interactive visualization tool, to reveal biases in neural networks.

    Prikaži ovu nit
    Poništi
  12. 13. sij

    Happy our new paper "Revealing Neural Network Bias to Non-Experts Through Interactive Counterfactual Examples" w/ , , , , and is now on arXiv:

    Prikaži ovu nit
    Poništi
  13. 11. sij

    Is it worth arguing about what deep learning is? Hasn't DL now become more of a buzzword than anything else? Maybe time for more precise definitions and fewer umbrella terms that encapsulate a myriad of different methods?

    Poništi
  14. 10. sij
    Poništi
  15. proslijedio/la je Tweet
    9. sij

    This year, the summer school is organized under the aegis of , our game AI research startup, and we have speakers from sharing their expertise, with more leading companies to be announced.

    Prikaži ovu nit
    Poništi
  16. 7. sij

    it's interesting to see what kind of representations emerge that the model is not directly optimized for. In this case, the network learned to predict events important for the survival of the agent (such as not being hit by a fireball). Code will follow soon.

    Prikaži ovu nit
    Poništi
  17. 7. sij

    Interestingly, the evolved forward model seems to have learned to predict if a fireball would hit the agent at the current position. Similarly to the great work by , , on observational dropout (),

    Prikaži ovu nit
    Poništi
  18. 7. sij

    As can be seen in the t-SNE plots in the video, the compressed information that is fed into the controller of the world model contains enough information to clearly separate environmental states in which the agent should move left or right.

    Prikaži ovu nit
    Poništi
  19. 7. sij

    This idea is very much inspired by Nick Cheney, , , great work on co-evolving morphologies and controllers for virtual creatures

    Prikaži ovu nit
    Poništi
  20. 7. sij

    The age objective counts how many generations passed since a mutation. Through a multiobjective EA, we optimize for both low age and high task reward.

    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:

    ·