Ted Benson

@edwardbenson

Head of Machine Intelligence at . Prior founder / YC Alum / PhD.

San Francisco, CA
Vrijeme pridruživanja: rujan 2008.

Tweetovi

Blokirali ste korisnika/cu @edwardbenson

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

  1. Prikvačeni tweet
    1. svi 2018.

    Come work with me! is growing an amazing team to bridge research and practice. - Machine learning engineers (NLP, computer vision, OCR) - UI / UX, HCI for data science / ML applications DM me for details. Offices in SF and NYC.

    Poništi
  2. proslijedio/la je Tweet
    2. velj

    Performance artist generates virtual traffic jams in Google Maps by pulling a wagon full of smartphones

    Poništi
  3. proslijedio/la je Tweet

    The WHO cannot share life saving information with Taiwan. Think about the absolute lunacy of this situation.

    Poništi
  4. proslijedio/la je Tweet

    The WHO “cannot share information with Taiwan... because Taiwan is not a member of the United Nations.”

    Poništi
  5. proslijedio/la je Tweet

    Breaking - Japanese Prime Minister Shinzo Abe said earlier today that it is necessary to let join the , otherwise it'll be very hard to contain the outbreak.

    Prikaži ovu nit
    Poništi
  6. proslijedio/la je Tweet

    The United Nation's plays a valuable role in ensuring aviation security. But silencing voices that oppose ICAO’s exclusion of Taiwan goes against their stated principles of fairness, inclusion, and transparency.

    Poništi
  7. proslijedio/la je Tweet
    28. sij

    is making complex data usable through a platform of connected black, white, and gray tools.“

    Poništi
  8. proslijedio/la je Tweet
    28. sij

    Dear , : You're communications officers for . Do you subscribe to this statement that your office gave to - which implies journalists, researchers, & US congressional staffers are lying about 's exclusion from your org?

    Prikaži ovu nit
    Poništi
  9. proslijedio/la je Tweet
    28. sij

    It's the moment we've all been waiting for! This month's release of nteract is out, including the long awaited nteract desktop app release. It includes: - support for ipywidgets - undo cell deletions - and more! Learn more here! ⬇️✨

    Poništi
  10. 25. sij

    “So if we skim one trillionth of a penny off of every Wasabi Roll sale, after seven lifetimes....”

    Poništi
  11. proslijedio/la je Tweet
    22. sij

    I removed a sentence from our 6-page paper being written in LaTeX. So now it is 6.5 pages.

    Prikaži ovu nit
    Poništi
  12. proslijedio/la je Tweet
    17. sij

    So proud of my wife today. Last year she quit her job to chase her dream of helping other mothers navigate the first few years of having children. After months of prep, she’s launching this week in SF! Info: Insta:

    Poništi
  13. 16. sij

    Two year ago, I ended a sales call with "we're not a good fit for you" Just sent a followup: "we're still not a good fit, but holy moly did I just meet the perfect researcher for you to fund if this problem still exists, 2 years on" I love this kind of long-game matchmaking

    Poništi
  14. 16. sij

    Apple’s UX moat is so wide that they literally shipped their flagship line of laptops with faulty keyboards for three years with zero impact on their install base. (Now we all just own external keyboards)

    Poništi
  15. 15. sij

    So back to the initial statement: The hand wavy claim is that “the big breakthrough” is likely to be the one that finds a way to jointly learn the representational space (and its axioms) alongside the embedding of instances and patterns that fill that space. 8/8

    Prikaži ovu nit
    Poništi
  16. 15. sij

    But fundamentally, a hierarchy is just another fixed geometric space with its own intrinsic long v short range problems We’ve traded a linear model for a more performing hierarchical one, but are still in affixing some pretty rigid constraints on what’s possible to represent 7/

    Prikaži ovu nit
    Poništi
  17. 15. sij

    We do this because we suspect human language is fundamentally hierarchical in structure; but also, honestly, because it makes (1) the math convenient and (2) produces results that are computationally tractable. 6/

    Prikaži ovu nit
    Poništi
  18. 15. sij

    Again, drawing on NLP, hierarchical modeling is a first order example of this. We fold space into a hierarchy such that long range dependencies from a linear perspective become close in a hierarchical space. 5/

    Prikaži ovu nit
    Poništi
  19. 15. sij

    In the same way that sfi-fi blockbusters are fond of scenes with a crazed scientist telling the protagonist the shortest distance between two points isn’t a line, it’s a fold of space, our AI models must learn to “fold space” — conceptually. 4/

    Prikaži ovu nit
    Poništi
  20. 15. sij

    Take natural language processing. One of the hardest problems is the complexity blowout of long-range dependency tracking. But the “long range-ed-ness” isn’t a given. It’s an artifact our difficulty imagining non-linear models of language and translating them into math. 3/

    Prikaži ovu nit
    Poništi
  21. 15. sij

    Or, another way to put it: that which brings us strong AI will probably be simple in construction but hard to think of in the first place.. if you think like Euclid: using our experiences in continuous, regularly metered space to mandate models of similar construction. 2/

    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:

    ·