Greg BrockmanVerified account

@gdb

Co-Founder & CTO

gregbrockman.com
Joined July 2010
106 Photos and videos Photos and videos

Tweets

You blocked @gdb

Are you sure you want to view these Tweets? Viewing Tweets won't unblock @gdb

  1. Mar 6

    Cool writeup in about the technique: "Shark or Baseball? Inside the ‘Black Box’ of a Neural Network"

    1 reply 11 retweets 43 likes
    Show this thread
    Show this thread
    Undo
  2. Mar 6

    A new technique for analyzing how neural nets make decisions. Not necessarily incomprehensible black boxes after all!

    0:52
    In collaboration with Google, we're releasing Activation Atlases: a new technique for visualizing what interactions between neurons can represent. 💻Blog: 📝Paper: 🔤Code: 🗺️Demo:
    4 replies 73 retweets 313 likes
    Show this thread
    Show this thread
    Undo
  3. Retweeted
    Mar 4

    The OpenAI policy and communications team are looking for a project manager! Very valuable position for those with project management experience to help us run things like internal groups, blog posts, and workshops.

    6 retweets 24 likes
    Undo
  4. Mar 4

    Excited to compete in my first strength competition on March 16th in Santa Clara, in an expo with over 3k athletes and 10k audience members: ! Been training hard, just like our reinforcement learning agents.

    3 replies 1 retweet 40 likes
    Undo
  5. Mar 4

    It’s never been more fun to be a reinforcement learning agent.

    We’re releasing a Neural MMO — a massively multiagent game environment that supports numerous populations of agents: - Code: - 3D Client:
    2 replies 29 retweets 216 likes
    Undo
  6. Retweeted
    Mar 3

    responds to ’s call for to open source the by insisting that "considering the impact and misuse of released models is the only sustainable path to progress in AI research."

    1 reply 14 retweets 28 likes
    Undo
  7. Feb 27

    Amazing to see OpenAI featured twice on 's 2019 list of Breakthrough Technologies (once for robotics, once for language): . Congrats to the team!

    2 replies 21 retweets 140 likes
    Undo
  8. Retweeted
    Feb 26

    What we learned hosting our first Spinning Up in Deep RL workshop: Plus: - Video lectures: - Spinning Up GitHub repo:

    5 replies 162 retweets 577 likes
    Undo
  9. Feb 25

    One very important aspect of OpenAI culture!

    The aspect of I've been most impressed with since joining is the aura of cooperation. I've seen very few cases of people prioritizing themselves before others, but many cases of people helping with each other's project without obvious personal gain
    8 retweets 67 likes
    Undo
  10. Retweeted
    Feb 22

    Now on the one hand, it *is* possible to bid for status by falsely claiming your work is dangerous. But it was obvious what GPT-2 could potentially wreck and how!

    2 replies 1 retweet 8 likes
    Show this thread
    Show this thread
    Undo
  11. Retweeted
    Feb 22

    One of the things that concerns me about the backlash to OpenAI’s responsible publication is that so much of it seems to be opposition to the very idea of responsible publication in AI.

    After withheld its latest algorithm, concerned that the research could have malicious uses, the AI community questioning whether it’s right to limit research in this way. I tried to capture some of the debate here:
    Show this thread
    4 replies 17 retweets 43 likes
    Undo
  12. Retweeted
    Feb 20

    Thread about my naive stance on the language model controversy. Forming an opinion on this now might be obsolete because things have moved quite fast these past two days. But I work on generative models and I wanted to take my time to think about this. 1/8

    3 replies 18 retweets 120 likes
    Show this thread
    Show this thread
    Undo
  13. Feb 20

    Enjoyed watching last night's conversation with . Very glad to see productive debate about publishing for GPT-2 and future powerful models (in a more nuanced way than is possible on Twitter):

    Tonight we held the 1st ever TWiML Live, an engaging discussion on the controversy behind the release of 's new language model. To learn more about this model, the concerns raised by its release, and how we can address them, don't miss the replay at
    2 replies 13 retweets 55 likes
    Undo
  14. Feb 19

    If you’re a social scientist, please consider applying to OpenAI!

    AI safety needs social scientists (and we're hiring!):
    12 retweets 88 likes
    Undo
  15. Retweeted
    Feb 18

    This is (by far) the most informed take I've read on the ML issues around the language modeling project yet. Almost none of the writing I've seen so far has been from people that have actually done language modeling research, so this is a breath of fresh air! :)

    I've finally decided to start a blog. I've been a bit frustrated with the Twitter conversation about 's GPT-2, and I haven't found a post that says all the things I want to say. So I wrote about it here: .
    Show this thread
    2 replies 68 retweets 303 likes
    Undo
  16. Retweeted
    Feb 17

    So, I've finally organized my thoughts surrounding 's decision to not release weights. I really empathize because I went through exactly this with SNAP_R research three years ago. Thread incoming. (1/12)

    3 replies 65 retweets 181 likes
    Show this thread
    Show this thread
    Undo
  17. Retweeted
    Feb 17

    The impact on society of this AI approach is impossible to predict right now

    An OpenAI employee printed out this AI-written sample and posted it by the recycling bin:
    19 replies 66 retweets 303 likes
    Undo
  18. Retweeted
    Feb 16

    I'd like to weigh in on the discussion. The decision not to release the trained model was carefully considered and important for norm-forming. Serving the public good requires us to draw lines on release somewhere: better long before catastrophe than after.

    12 replies 113 retweets 396 likes
    Show this thread
    Show this thread
    Undo
  19. Retweeted
    Feb 16

    I'm surprised that so many are opposed to the idea that, as AI becomes increasingly capable, publication norms in AI should change.

    29 replies 48 retweets 416 likes
    Undo
  20. Retweeted
    Feb 14

    Here's a ridiculous result from the GPT-2 paper (Table 13) that might get buried --- the model makes up an entire, coherent news article about TALKING UNICORNS, given only 2 sentences of context. WHAT??!!

    41 replies 549 retweets 1,172 likes
    Show this thread
    Show this thread
    Undo

@gdb hasn't Tweeted yet.

Loading seems to be taking a while.

Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.

    You may also like

    ·