David Fendrich

@fnedrik

Programmer, data scientist, AI aficionado. CTO of . Plays too much chess.

Gothenburg
Vrijeme pridruživanja: travanj 2008.

Tweetovi

Blokirali ste korisnika/cu @fnedrik

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

  1. 30. sij

    I just published a post on which machine learning methods we use (for our diverse set of projects and customers) in practice at Tenfifty:

    Poništi
  2. proslijedio/la je Tweet
    7. sij

    That GPT-2 can learn to play chess just from reading notation (no knowledge of the game itself) seems to suggest that these DL systems do figure out useful representations on their own. If you can fake task X well enough you have understood *something*.

    Poništi
  3. proslijedio/la je Tweet

    One of my very favorite space images of all time: snowstorm on a comet (plus a whole lotta cosmic ray hits to the sensor). May all your solstices include reflections on the beauty and depth of the exquisite universe around us!

    Poništi
  4. proslijedio/la je Tweet
    13. stu 2019.

    (3/5) After a day of training, he can withstand significant perturbations (but sometimes you just get too many cubes in the face…):

    Prikaži ovu nit
    Poništi
  5. proslijedio/la je Tweet
    9. stu 2019.

    Some of the most interesting mimicry I’ve ever seen — the wings of this moth (Macrocilix maia) mimic two flies feasting on bird droppings.

    Prikaži ovu nit
    Poništi
  6. proslijedio/la je Tweet
    25. lis 2019.

    how to hide your Bitcoins

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

    It happened! XENOMORPH ALIEN! Fully freehand , no pattern! He glows in the dark too!!

    Prikaži ovu nit
    Poništi
  8. proslijedio/la je Tweet

    Black Mirror is entirely superfluous.

    Prikaži ovu nit
    Poništi
  9. proslijedio/la je Tweet
    21. ruj 2019.

    A brilliant quote crossed my desk which I could not let go un-shared (from Haldane JBS. , Science and life , 1969): "I consider it desirable that a man's or a woman's major research work should be on a subject in which he or she has not taken a degree."

    Poništi
  10. proslijedio/la je Tweet

    Today is climate strike day. Millions of people across the world, according to the organisers, will be marching to save the planet. But having dug into what they're actually marching for, I'm profoundly uneasy. So uneasy, in fact, that it's time for a THREAD.

    Prikaži ovu nit
    Poništi
  11. proslijedio/la je Tweet
    29. kol 2019.

    THREAD on a wacky problem that AI ethics theorists should be aware of. Consider a coin flip bet: Heads you get 1.5 times your current wealth. Tails you get 0.6. Should you take it? Expected value is 1/2*0.6 + 1/2*1.5 = 1.05 times your wealth. So take it right? Not quite...

    Prikaži ovu nit
    Poništi
  12. proslijedio/la je Tweet
    19. kol 2019.

    Hur levererar man på topp när man ska hålla föredrag? Tips från Freud.

    Poništi
  13. proslijedio/la je Tweet
    27. srp 2019.

    Here’s a cool instance of the Banach fixed-point theorem

    Poništi
  14. proslijedio/la je Tweet
    17. srp 2019.

    The wonder of surveys, an example. Would you murder your mother for $5? - 73% no - 27% yes Of those 27%: - 9% misunderstood the question - 4% attempted "no", but had an input error - 12% are trolling - 2% are genuine

    Prikaži ovu nit
    Poništi
  15. proslijedio/la je Tweet

    Nearly 20,000 inhabitants of a Turkish province called ‘Batman’ have signed a petition to re-shape the province’s borders into the shape of the Bat-Signal.

    Poništi
  16. proslijedio/la je Tweet

    We've all been targeting 10k steps a day because some Japanese pedometer company thought the character for 10,000 (万) looks like a person walking...

    Prikaži ovu nit
    Poništi
  17. proslijedio/la je Tweet
    6. lip 2019.

    A brilliant tuned mass damper demonstration with simple paper [original source, Luca Bellini: ] [more on mass dampers: ]

    Poništi
  18. 14. svi 2019.

    You may know a well calibrated percentage only by its inevitable betrayal.

    Poništi
  19. proslijedio/la je Tweet
    8. svi 2019.

    How would I know if my own research area was this wrong? Our usual safeguards won’t save us: peer review, meta-analysis, 100s of conceptual replications, listening to eminent researchers. All failed. This should be keeping us up at night.

    Poništi
  20. proslijedio/la je Tweet
    6. svi 2019.

    nice paper showing neural net embedding methods aren't actually better than SVD based embedding methods:

    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:

    ·