Andrej KarpathyVerified account

@karpathy

Research scientist at OpenAI. Previously CS PhD student at Stanford. I like to train Deep Neural Nets on large datasets. On a quest to solve intelligence.

Stanford
cs.stanford.edu/~karpathy/
Joined April 2009
378 Photos and videos Photos and videos

Tweets

@karpathy is blocked

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

  1. Jan 23

    It took 50 years for the world to install the first million industrial robots. The next million will take only eight

    1 reply 128 retweets 167 likes
  2. Jan 22

    Aww the Google Self Driving Car project is already part of the Computer _History_ Museum?

    6 replies 60 retweets 188 likes
  3. Jan 20

    Everything I know about design of ConvNets (resnets, bigger=better, batchnorms etc) is useless in RL. Superbasic 4-layer ConvNets work best.

    18 replies 189 retweets 603 likes
  4. Jan 18

    Dear , I got the 100 mails with your AMAZING credit card offer. They aren't just getting lost. Please stop the torture.

    7 replies 6 retweets 72 likes
  5. Jan 18

    w00t MIT's Deep Learning for Self-Driving Cars class uses ConvNetJS. DeepTraffic: &DeepTesla

    4 replies 134 retweets 323 likes
  6. Jan 18

    Imperative, dynamic graph construction is going strong recently, also with recent & v nice looking minpy , DyNet, etc

    2 replies 35 retweets 126 likes
  7. Jan 18

    Excited to see PyTorch (a new Deep Learning library) released! Tried it for few days, it is awesome: imperative!, fast, clean and simple.

    GPU Tensors, Dynamic Neural Networks and deep Python integration. Hello world!
    8 replies 278 retweets 561 likes
  8. Jan 17

    "Personally, I do not trust paper results at all. I tend to read papers for inspiration" A correct rant.

    14 replies 240 retweets 547 likes
  9. Jan 17

    Wrote up some thoughts on VR (long interest of mine) in a blog post: "Virtual Reality: still not quite there, again"

    10 replies 64 retweets 155 likes
  10. Jan 15

    I don't understand why Earth over last few B years was not an easy target for an alien superintelligence when galaxy is only ~100k LY across

    57 replies 15 retweets 94 likes
  11. Jan 13

    Last day for applications to Residency for 2017

    Just seven months into the Google Brain Residency program, our first group of residents are already making an impact
    1 reply 17 retweets 44 likes
  12. Jan 13

    I sequestered myself in a conference room last week (was ill) instead of open seating & RescueTime shows 1.8x more productivity. Interesting

    9 replies 17 retweets 129 likes
  13. Jan 13

    Greg's post on past/present/future of OpenAI including fun stories of OpenAI early days

    84 retweets 211 likes
  14. Jan 11

    GTA V + Universe:

    15 replies 398 retweets 684 likes
  15. Jan 10

    Not clear why TF still really likes `reduce_` syntax bloat, or `keep_dims` vs numpy's `keepdims`, etc.

    11 replies 7 retweets 46 likes
  16. Jan 10

    TensorFlow 1.0.0-alpha many numpy API compatibility changes are very welcome, seems could still go even further

    1 reply 163 retweets 269 likes
  17. Jan 7

    TV anchor says "Alexa order me a dollhouse" on live TV, Alexas in people's homes activate and go on shopping spree

    7 replies 180 retweets 262 likes
  18. Jan 2

    A tech report summarizing my NIPS tutorial on GANs

    9 replies 338 retweets 695 likes
  19. 30 Dec 2016

    Local news from Mission SF each time: "[something terrible] @ [a place I walk by all the time]. no arrests made."

    5 replies 3 retweets 20 likes
  20. 24 Dec 2016

    Fun fact 27/120: There are nuclear submarines out there carrying 40 nuclear warheads controlled by a computer running Windows XP.

    20 replies 160 retweets 298 likes

@karpathy 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

    ·