Jérémy Perret

@gyrodiot

R&D at - All opinions personal, I do not speak for my employer. I promise I'll post the purpose of my life once I figure it out.

Vrijeme pridruživanja: travanj 2010.
Rođen/a 27. veljače 1992.

Tweetovi

Blokirali ste korisnika/cu @gyrodiot

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

  1. Prikvačeni tweet
    20. srp 2018.

    I now have a French-speaking account, where I will rant about and various topics, avoiding English references as much as possible. I'll cross-post important things. Thanks for your attention!

    Poništi
  2. 26. sij

    The complex mesh of heuristics and embeddings and clusterings (the « YT algorithm ») is not monolithic. It is designed to offload many decisions to users, creators and a few curators. It leverages our power. There is no simulation, no self-play. Only our feedback, for now. (3/3)

    Prikaži ovu nit
    Poništi
  3. 26. sij

    ... the stated objective of YT search & discovery (« improving the experience »), yet doesn't discuss ads on monetization at all. It does mention mental health and politics a bit. The concept I found most interesting was: we are a major part of the algorithm. We sort stuff! (2/3)

    Prikaži ovu nit
    Poništi
  4. 26. sij

    Just finished this episode of 's AI Podcast . Mostly watched it so I could know if it was worthy of a recommendation to . And, yes it is. The conversation does talk about the rough criteria for YT recommendations... (1/3)

    Prikaži ovu nit
    Poništi
  5. proslijedio/la je Tweet
    31. pro 2019.

    "It means it is possible to run without being able to walk. If the canons of psycholinguistics say this is impossible, so much the worse for them, and so much the better for our understanding of the human brain."

    Poništi
  6. 29. pro 2019.

    Maybe I should have chosen a narrow topic I can talk hours about, and state a hundred things about it, instead of trying to generate fresh insights on the fly. I will try again.

    Poništi
  7. 18. pro 2019.

    6. If you spend time optimizing a process for yourself, more time that is saved doing so, and you cannot save the time of others sharing your solution, this means you're now doing a very unique thing, very well. Which means it has either a lot of value, or none at all.

    Prikaži ovu nit
    Poništi
  8. 18. pro 2019.

    5. The goal of automation is to bridge the gap between « I make this happen reliably » and « This happens reliably ». Fighting constantly for the same outcome is a tragedy. When you understand a method well enough, the next step is to never apply it again.

    Prikaži ovu nit
    Poništi
  9. 17. pro 2019.

    4. « Any proposal that begins “if everyone would just…”, is already doomed. Everyone will not just. Never in history has everyone just. » (quoted from ). Collective action problems are hard, and worth studying.

    Prikaži ovu nit
    Poništi
  10. 16. pro 2019.

    3. You cannot expect a clean definition of artificial intelligence until you have one of intelligence itself. You also don't need either to create it. Our biology didn't come with a manual, yet here we are.

    Prikaži ovu nit
    Poništi
  11. 15. pro 2019.

    2. When you try to make machines do what they can't do yet, you always run into the same sub-problems, which have been studied for decades. Learn your AI history by exploring naive and unexpectedly hard questions. You will not be the first and that's OK.

    Prikaži ovu nit
    Poništi
  12. 15. pro 2019.

    1. The edge of a field is noisy. Do not bother with the state of the art, stay a few weeks behind, use the saved time to consolidate what you know. Even if your job requires you to know the edge, it's more valuable to know how it expands than to know where it is.

    Prikaži ovu nit
    Poništi
  13. 15. pro 2019.

    Let's do this. 3 opinions per like, as quickly as I can answer, on my favorite subjects by default. If you want a particular topic, please reply.

    Prikaži ovu nit
    Poništi
  14. 5. stu 2019.

    That's what I've got so far. To reiterate, these are my personal parameters, instructions for my future self. I expect they will continue to be tweaked. I came up with them through experimentation, logging, and patient self-reflection. I'm happy with this investment. (6/6)

    Prikaži ovu nit
    Poništi
  15. 5. stu 2019.

    Fourth: outlets. The optimal state will continue as long as you have somewhere to express your ideas. Active memorization is the death of attention. Write/draw stuff, discuss your ideas with people. If you find yourself out of working memory, game over. (5/6)

    Prikaži ovu nit
    Poništi
  16. 5. stu 2019.

    Third: coffee. A small dose works exceedingly well to reach the optimal state but will not, ever, solve waking-up brain fog. Do not waste precious coffee early in the morning. Drink water, have a walk, some fresh air, smell some flowers instead. (4/6)

    Prikaži ovu nit
    Poništi
  17. 5. stu 2019.

    Second: hunger-based food intake. Have healthy snacks always ready, eat whenever you're hungry, stop whenever you aren't anymore. Overeating means brain fog. Proper, nice full meals belong to the evening, unless you want to work late (even then: do not resist sleep!) (3/6)

    Prikaži ovu nit
    Poništi
  18. 5. stu 2019.

    First and foremost: sleep. A full night, then wait a few hours. Naps won't do, brain stays foggy right after waking up. Sleep deprivation is guaranteed brain fog. Do not resist sleep in the evening. If you have to, wake up earlier, but go. To. Sleep. (2/6)

    Prikaži ovu nit
    Poništi
  19. 5. stu 2019.

    Quick personal thread. I've been trying to figure out for years a reliable formula to reach my optimal mental state, full creativity/attention, flow, however you name it. I've narrowed it down to a few instructions, which work for me, probably not for you. (1/6)

    Prikaži ovu nit
    Poništi
  20. 9. lis 2019.

    Like last year, mention of the UN Sustainable Development Goals The 17 goals are a call to action, which is fitting for an AI-for-good session. Everyone will claim they're doing good though. We have to ensure it's done in an effective way.

    Prikaži ovu nit
    Poništi
  21. 9. lis 2019.

    This model for the future of work here, with forecasting, cooperation & resource optimization looks desirable however hard job automation hits. It's excellent. Not a contingency plan, but a bright vision. I'd be annoyed if we waited for a crisis to implement it.

    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:

    ·