Danny Hernandez

@Hernandez_Danny

Measuring and forecasting AI progress

San Francisco, CA
Vrijeme pridruživanja: ožujak 2011.

Tweetovi

Blokirali ste korisnika/cu @Hernandez_Danny

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

  1. Prikvačeni tweet
    16. svi 2018.

    I'm really proud to bring some empirical evidence to the what kind of AI progress should we expect conversation. I expect increasing compute to unlock new capabilities in the near future.

    Poništi
  2. proslijedio/la je Tweet
    30. sij

    We're standardizing OpenAI's deep learning framework on PyTorch to increase our research productivity at scale on GPUs (and have just released a PyTorch version of Spinning Up in Deep RL):

    Poništi
  3. 21. sij

    Therapists also prescribe drugs, which seems important. But you don't need to see someone weekly to get prescribed drugs.

    Prikaži ovu nit
    Poništi
  4. 21. sij

    Therapists are generally reluctant to give advice. It was probably intractable to teach such a large profession to give good advice broadly, but medical-grade listening skills was a viable core product.

    Prikaži ovu nit
    Poništi
  5. 21. sij

    The market price for trained listening seems to be ~$200/hr. (therapists) Seems worth cultivating listening skills, as most people value listening highly and will never seek out therapists.

    Prikaži ovu nit
    Poništi
  6. 20. sij

    People don't want to feel small, and it's generally not advantageous to make them feel that way. The virtue of humility could also be defined as intentionally underestimating yourself, but an accurate picture of your abilities seems like a better thing to strive for.

    Prikaži ovu nit
    Poništi
  7. 19. sij

    The aspect of humility worth cultivating is avoiding making others feel small.

    Prikaži ovu nit
    Poništi
  8. 10. sij

    6. If there is an AI Autumn it will look like increased focus on lucrative applications. For these reasons I consider AI/ML as a relatively safe career bet. AI/ML is relevant to a core or the core problem for most big Tech companies. — end of nuance-mongering —

    Prikaži ovu nit
    Poništi
  9. 10. sij
    Prikaži ovu nit
    Poništi
  10. 10. sij

    4. RL is powering Facebook notifications. RL isn't considered to just work the way supervised learning often does, so I find this pretty meaningful. Facebook as a business deeply cares about notifications.

    Prikaži ovu nit
    Poništi
  11. 10. sij

    3. The biggest upgrade to Google in years leveraged BERT/Transformers/Deep Learning.

    Prikaži ovu nit
    Poništi
  12. 10. sij

    2. In the AI media hype wars the possibility of an AI winter is often raised, but in a way that feels misleading/ill-defined. Past AI winters consisted of huge numbers of companies folding and drastic cuts to funding.

    Prikaži ovu nit
    Poništi
  13. 10. sij

    AI Winter is NOT coming 🤖❄️ Maybe there will be an AI Autumn — Begin nuance-mongering — A Winter of old would look like FaceGooZon laying off more than 30% of it's AI/ML staff. I'd put less than 1% on that in next 3 years. AI/ML is producing too much business value (1 of 6)

    Prikaži ovu nit
    Poništi
  14. proslijedio/la je Tweet
    8. sij

    Twitter will add 4 settings on who can reply to your tweets. • Global lets anyone reply • Group is for people you follow & mention • Panel is people you specifically mention in the tweet • Statement allows you to post a tweet and no one can reply

    Prikaži ovu nit
    Poništi
  15. proslijedio/la je Tweet
    1. sij

    Write down your 2020 goals and look at them on the 1st of every month. It's worth it!

    Poništi
  16. proslijedio/la je Tweet
    7. pro 2019.

    The Lesson to Unlearn:

    Poništi
  17. proslijedio/la je Tweet
    3. pro 2019.

    We're releasing Procgen Benchmark, 16 procedurally-generated environments for measuring how quickly a reinforcement learning agent learns generalizable skills. This has become the standard research platform used by the OpenAI RL team:

    Poništi
  18. proslijedio/la je Tweet
    23. stu 2019.

    The Bus Ticket Theory of Genius:

    Poništi
  19. proslijedio/la je Tweet
    21. stu 2019.

    We're releasing Safety Gym, environments and tools to evaluate reinforcement learning with safety constraints: Aims to ultimately help agents satisfy real-world safety requirements while training (eg not driving off a cliff, not writing abusive content).

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

    When people visit your startup, they should be surprised how few people you have. A visitor who walks around and is impressed by the magnitude of your operation is implicitly saying "Did it really take all these people to make that crappy product?"

    Poništi
  21. 7. stu 2019.

    I think it's super interesting that past AI winters don't even register when you zoom out and look at compute trends over the last 60 years.

    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:

    ·