Jessy Lin

@realJessyLin

CS, AI, philosophy, and other fun things @ MIT

Vrijeme pridruživanja: ožujak 2013.

Tweetovi

Blokirali ste korisnika/cu @realJessyLin

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

  1. 11. sij

    Reading abt how binary was just an “arcane form of arithmetic” when Shannon realized connection to logic. Imagine how ridiculous it seemed - “so we should adopt this silly form of adding numbers over how humans have counted for millennia? nice” What are equivalent ideas today?

    Poništi
  2. 21. pro 2019.

    no matter how many corrective measures we implement on top, the current machine learning paradigm is fundamentally based on the assumption that the future will be more of the same. for all their biases / irrationalities, humans realize this doesn't have to be true.

    Prikaži ovu nit
    Poništi
  3. 21. pro 2019.

    the common thread between most unethical or questionable use cases for AI (e.g. predictive policing, risk assessment): the creators have failed to ask "can the future be different from the past?"

    Prikaži ovu nit
    Poništi
  4. 25. stu 2019.

    new favorite icebreaker question (I o/w dread the unoriginal "spirit animal" or "favorite hobby" stuff): "what was your last Amazon purchase?" makes for such good stories / actual bonding!

    Poništi
  5. proslijedio/la je Tweet
    22. stu 2019.

    It's taken a "village", and over two years of work, but we've finally pushed through to share transparent details about how well models work, and the relationship to "ML fairness" criteria.

    Poništi
  6. 17. stu 2019.

    you know you're too deep into it when you search up "rich models" expecting statistical models but you get heidi klum instead

    Poništi
  7. proslijedio/la je Tweet
    6. stu 2019.

    Sometimes, when you meet a Nature editor at a conference and start ranting about theory and measurement, they ask you to write about it. Here's a very short thing I wrote.

    Poništi
  8. proslijedio/la je Tweet
    3. lis 2019.

    New essay, joint with : "How can we develop transformative tools for thought?"

    Poništi
  9. 2. lis 2019.

    from an analysis of ICML 2012 paper reviews: "The core difficulty is that the quality of the papers a reviewer receives varies substantially more than the reviewing tendencies of reviewers." (seems obvious in hindsight, but people rarely think about this)

    Prikaži ovu nit
    Poništi
  10. 2. lis 2019.

    interesting tidbit: given judges with different rating tendencies, we take it for granted that we can just normalize their scores to account for their biases, but this can _hurt_!

    Prikaži ovu nit
    Poništi
  11. 16. ruj 2019.

    i'm officially labeled a psycholinguist, do i get my phd now

    Poništi
  12. 2. svi 2019.
    Prikaži ovu nit
    Poništi
  13. 2. svi 2019.
    Prikaži ovu nit
    Poništi
  14. 2. svi 2019.
    Prikaži ovu nit
    Poništi
  15. 2. svi 2019.
    Prikaži ovu nit
    Poništi
  16. 8. tra 2019.

    Things like this make me so excited / frustrated at the same time about progress in understanding our own biology—feels like software's finally at the pt we can do sth, but we lack the hardware / measurement tools (instead we get Fitbit/"log how you feel")

    Poništi
  17. 14. ožu 2019.

    this was really cool — re: how we think when we interact with others. step 1 is finding values that guide the way we relate to people; step 2 is realizing your very belief in / adherence to those values is a policy itself

    Poništi
  18. 28. sij 2019.

    kind of interesting in general how it seems in fashion to make evolutionary arguments for everything -- so many things I'm recommended nowadays have this flavor of analysis

    Prikaži ovu nit
    Poništi
  19. 28. sij 2019.

    Along the lines of all these popular books (Sapiens, Elephant in the Brain, etc.) about how human nature manifests in politics / society, Bertrand Russell on the very same topic >60 years ago: (thanks !)

    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:

    ·