Phil Tully

@phtully

Brain-inspired ,

New York, NY
linkedin.com/in/phtully/
Vrijeme pridruživanja: listopad 2012.
1 fotografija ili videozapis Fotografije i videozapisi

Tweetovi

Blokirali ste korisnika/cu @phtully

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

  1. proslijedio/la je Tweet
    prije 22 sata

    We have developed a new technique to mark the images in a data set so that researchers can determine whether a particular machine learning model has been trained using those images. Learn more about “radioactive data” here.

    5 replies 200 proslijeđenih tweetova 621 korisnik označava da mu se sviđa
    Poništi
  2. proslijedio/la je Tweet
    4. velj

    Disinformation is one of the internet's most pressing challenges. We’re excited to share our latest research and experiments designed to thwart disinformation. Learn more about our work.

    13 proslijeđenih tweetova 20 korisnika označava da im se sviđa
    Poništi
  3. proslijedio/la je Tweet
    30. sij

    People of the Internet (esp. twitter), I am pleased to announce the launch of the at ! (just in time for !)

    1 reply 6 proslijeđenih tweetova 16 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  4. proslijedio/la je Tweet
    30. sij

    We’re turning on a tool for key moments of the 2020 US election that enables people to report misleading information about how to participate in an election or other civic event.

    178 replies 482 proslijeđena tweeta 658 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  5. 17. sij

    NEW: The Data Science team has two open summer '20 internship positions. interns are paid and will be expected to work out of our Reston, VA office. Come help us tackle problems at the intersection of machine learning and cybersecurity:

    10 proslijeđenih tweetova 13 korisnika označava da im se sviđa
    Poništi
  6. proslijedio/la je Tweet
    16. sij

    Introducing Reformer, an efficiency optimized architecture, based on the Transformer model for language understanding, that can handle context windows of up to 1 million words, all on a single accelerator with only 16GB of memory. Read all about it ↓

    14 replies 474 proslijeđena tweeta 1.354 korisnika označavaju da im se sviđa
    Poništi
  7. proslijedio/la je Tweet
    20. pro 2019.

    Looks like 1st significant eg of AI-generated faces in campaign. We'll need holistic collaboration on detection - not just whack-a-mole to try detect each indiv faked image (which will be hard)

    1 reply 5 proslijeđenih tweetova 14 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  8. proslijedio/la je Tweet
    11. stu 2019.

    can improve , but it requires substantial forms of control to be (rather than ), I argue with McCutcheon and in NATURE - Mach. Intell.

    1 reply 26 proslijeđenih tweetova 52 korisnika označavaju da im se sviđa
    Poništi
  9. proslijedio/la je Tweet
    3. pro 2019.

    Yes. This. We’ve already seen how copied-and-pasted text spread across multiple and diverse “news” creates a kind of astroturf that may trick our brains into thinking we’re triangulating. When that text can be auto-generated (and varied) that effect will be even more powerful.

    In the current state of disinformation and information operations, I see more harm coming from the potential of machine generated text applied to disinfo at scale than from other forms of synthetic media. Thanks for having me this morning, !
    8 replies 62 proslijeđena tweeta 164 korisnika označavaju da im se sviđa
    Poništi
  10. proslijedio/la je Tweet
    21. stu 2019.

    Here is dropping Malware Data Science knowledge Phil is the DS behind the FLARE joint Malware Analysis tool StringSifter

    1 reply 3 proslijeđena tweeta 22 korisnika označavaju da im se sviđa
    Poništi
  11. proslijedio/la je Tweet
    19. stu 2019.

    Experts in online terrorism and trolling are starting to explore the danger that improved language software could empower bad actors. Their experiments involve testing how well software can mimic or amplify online disinformation and propaganda material.

    23 proslijeđena tweeta 41 korisnik označava da mu se sviđa
    Poništi
  12. 14. stu 2019.

    Proud of the results achieved by my data science intern over this past summer. We learned a lot working closely with the experts on 's IO Analysis team - watch this space for more cool ML approaches to these problems. Great job Sajid!

    BLOG | By fine-tuning an open source Transformer known as GPT-2, we can detect social media posts being leveraged in despite their syntactic differences to the model’s original training data. >> Read more about our findings:
    7 proslijeđenih tweetova 22 korisnika označavaju da im se sviđa
    Poništi
  13. proslijedio/la je Tweet
    8. stu 2019.

    Transformers are made of interleaved self-attention and feedforward sublayers. Can we find a better pattern? New work on *improving transformers by reordering their sublayers* with and

    7 replies 79 proslijeđenih tweetova 323 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  14. proslijedio/la je Tweet
    7. stu 2019.

    Our new paper: Unsupervised Cross-lingual Representation Learning at Scale We release XLM-R, a Transformer MLM trained in 100 langs on 2.5 TB of text data. Double digit gains on XLU benchmarks + strong per-language performance (~XLNet on GLUE). [1/6]

    5 replies 148 proslijeđenih tweetova 426 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  15. proslijedio/la je Tweet
    5. stu 2019.

    We're releasing the 1.5billion parameter GPT-2 model as part of our staged release publication strategy. - GPT-2 output detection model: - Research from partners on potential malicious uses: - More details:

    60 replies 706 proslijeđenih tweetova 1.792 korisnika označavaju da im se sviđa
    Poništi
  16. proslijedio/la je Tweet
    31. lis 2019.

    Today we are very proud to open-source Spleeter a music source separation tool made here at R&D. Part of the team will be at to demonstrate it's capabilities. See you in Delft if you plan to attend and happy hacking everyone!

    33 proslijeđena tweeta 69 korisnika označava da im se sviđa
    Poništi
  17. proslijedio/la je Tweet
    31. lis 2019.

    Today is publicly unveiling , newest espionage dataminer that we've observed being used at a telecom providers.

    136 proslijeđenih tweetova 210 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  18. proslijedio/la je Tweet
    29. lis 2019.

    Summarization is one of the most important & least solved tasks in Problem with all models: they are not optimized for factual correctness We introduce a new task, dataset and model. Work by Paper

    243 proslijeđena tweeta 703 korisnika označavaju da im se sviđa
    Poništi
  19. proslijedio/la je Tweet
    29. lis 2019.

    Check out the now available SecML v0.10 with and library connector (beta). In addition, multiple enhancements to support. New tutorials available here: Code: Changelog:

    , , i još njih 2
    8 proslijeđenih tweetova 15 korisnika označava da im se sviđa
    Poništi
  20. 28. lis 2019.

    Reminder that is hiring motivated Data Scientists to work on similar problems at the intersection of machine learning and cybersecurity. Roles available at various experience levels, and remote work is an option. Reach out for details or see:

    data scientist presenting an ML approach for ranking strings output during binary analysis with StringSifter. Available at: .
    1 proslijeđeni tweet 8 korisnika označava da im se sviđa
    Poništi

@phtully 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:

    ·