Felix Hill

@FelixHill84

Research Scientist, Deepmind. Language, grounding, psychology, abstraction, analogy, reasoning, 90s football, 80s music. Opinions my own

London, England
fh295.github.io
Vrijeme pridruživanja: srpanj 2010.
23 fotografije ili videozapisa Fotografije i videozapisi

Tweetovi

Blokirali ste korisnika/cu @FelixHill84

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @FelixHill84

  1. Prikvačeni tweet
    10. sij

    So this year, I'm starting a Twitter-friendly micro-blog. I'll Tweet each entry (# 1 coming right up), try to do one every couple of weeks, and record them all on the blog site (). It's called *NonCompositional*

    7 replies 11 proslijeđenih tweetova 64 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  2. prije 2 sata

    Unfortunately Lego had also been in the news recently for less positive reasons: promoting an unrealistic body image to kids.

    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  3. prije 2 sata

    Did you know: language change like this is often driven by migration and meetings of different mother tongues. and I, and other fantastic collabs did. Lego is particularly interesting as a loanword for children all over the world.

    1 reply
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  4. prije 2 sata

    Settled by the court of Twitter and . If you have to count them, it's "Lego bricks" or "pieces of Lego". I presume the North American books that are cementing the 'Legos' thing will now be slowly removed from circulation.

    Odgovor korisniku/ci @FelixHill84
    In Danish LEGO is a brand (proper noun, not listed in official dictionary - ). LEGO bricks (da: legoklods) is what you count and step on.
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  5. prije 11 sati

    If you think *I emptied all of the Legos out onto the carpet* sounds like a reasonable sentence, you may be surprised to hear that at least 60m of us over here don't. Isn't language amazing.

    Blimey! Lego is a legitimate count noun in American English. See attached graphs. To me it will always be exclusively a mass noun. What will you guys do to our language next?!?!?! Perhaps someone Danish should settle this once and for all
    Prikaži ovu nit
    Poništi
  6. prije 12 sati

    THIS IS NOT A SERIOUS TWEET. Just saying.

    1 reply 2 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  7. prije 12 sati

    Blimey! Lego is a legitimate count noun in American English. See attached graphs. To me it will always be exclusively a mass noun. What will you guys do to our language next?!?!?! Perhaps someone Danish should settle this once and for all

    5 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  8. prije 14 sati

    Sweating the big stuff...

    2:18
    ICYMI: 1978: The British public were not impressed at the prospect of switching from miles to kilometres.
    1 korisnik označava da mu se sviđa
    Poništi
  9. 3. velj

    By forcing it beyond the sports pages, ruins my plan to avoid knowing the Superbowl result until I can watch on UK time. Take your geography classes kids!

    1 reply
    Poništi
  10. 3. velj

    I should have put that in quotations - not my words! Really clear and easy to digest review and outlook in PNAS by

    1 korisnik označava da mu se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  11. 3. velj

    Deep learning was inspired by the architecture of the cerebral cortex and insights ... general intelligence may be found in other brain regions that are essential for planning and survival, but major breakthroughs will be needed to achieve these goals.

    1 reply 4 proslijeđena tweeta 9 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  12. 3. velj

    Its complicated because there may be distinct ethical and/or practical issues in the natural and artificial cases. But from a purely scientific point of view I do think we may learn a bit more from lessons from the challenges of analysis in biology.

    coming from a biology background, I’ve often wondered this as well. Natural systems can be understood by observation, yet computational ones are supposed to be transparent.
    1 reply 2 korisnika označavaju da im se sviđa
    Poništi
  13. proslijedio/la je Tweet
    2. velj

    The McGurk Effect takes another victim

    11 replies 51 proslijeđeni tweet 124 korisnika označavaju da im se sviđa
    Poništi
  14. 2. velj

    Lots of connections to NALU and extensions, applied to free-form natural text

    "Giving BERT a calculator" Big step in adding computational abilities to language models. Still no understanding but can certainly start to fool you. What happens if you "give BERT" thousand other similar abilities? cc
    4 proslijeđena tweeta 12 korisnika označava da im se sviđa
    Poništi
  15. proslijedio/la je Tweet
    1. velj

    Now out:

    Very exited about this work led by Cory Shain and , also with and William Schuler: " fMRI reveals language-specific predictive coding during naturalistic sentence comprehension" 1/4
    Prikaži ovu nit
    3 proslijeđena tweeta 21 korisnik označava da mu se sviđa
    Poništi
  16. 1. velj

    Is this necessarily a bad thing? Should we expect that such complexity can ever be explained (nomologically) in terms that we can comprehend? 3/3

    15 replies 10 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  17. 1. velj

    Artificial systems without any learning component are well understood, by those who build them, but if we choose to do ML, surely we must commit to learning algorithms finding structures or functions in data that we wouldn't have conceived of? 2/N

    15 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  18. 1. velj

    I find the question of NNs and interpretability a curious one. What is an example of a complex system (natural or biological) that you think is 'sufficiently' understood, explained or interpretable? 1/N

    12 replies 11 proslijeđenih tweetova 91 korisnik označava da mu se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  19. 1. velj

    Looks like direct replication of a result we showed already, thanks to brilliant exp design by . A citation would have been nice tho!

    Odgovor korisnicima @hardmaru @kevin_zakka @togelius
    Yes, et. al's paper () supports this hypothesis too. Switching to an agent-centric view decreased performance on the training task, but led to better generalization on test tasks.
    4 korisnika označavaju da im se sviđa
    Poništi
  20. 31. sij

    Marvellous...

    So enamoured by Brexit they're willing to go to Downing Street to watch it happen on a big clock. Not able to list a single fucking benefit it'll have on our lives.
    6 korisnika označava da im se sviđa
    Poništi
  21. proslijedio/la je Tweet
    29. sij

    Joining on , lexical ambiguity from Rodd 2020 (): "more than 80% of word forms are polysemous—they can refer to more than one related word sense [...] often viewed within psycholinguistics as a troublesome nuisance"

    2 proslijeđena tweeta 8 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi

@FelixHill84 još nije objavio nijedan Tweet.

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:

    ·