How did Andrej Karpathy become a rockstar?

Oracle blahbl
Feb 14 51 Comments

His thesis is good, but not extraordinary. How did he have such a meteorite rise to become head of AI at Tesla just an year after phd?

comments

Want to comment? LOG IN or SIGN UP
TOP 51 Comments
  • Facebook hxeM71
    I guess you need to be socially famous. It's less about the actual PhD work.
    Feb 14 7
    • Oracle blahbl
      OP
      Interesting. Is that the way to achieve greatness in AI?
      Feb 14
    • Yahoo k_
      Yep
      Feb 14
    • Walmart.com / Eng äččä
      @blahbl I think that ship has sailed
      Feb 15
    • Oracle blahbl
      OP
      What do you mean? There’s always new stuff coming up like ai ethics etc. but yeah, they were in right place at the right time
      Feb 15
    • Walmart.com / Eng äččä
      I mean now even if some new trend comes in deep learning, it’s unlikely that you will become as famous as Karapathy by making videos and websites
      Feb 15
  • Google BeataMaria
    Meteorites don’t rise and corporate execs aren’t rockstars
    Feb 14 0
  • Walmart.com salamanca
    Stanford has extraordinary PR machinery. It also has deep tentacles into the bay area tech industry. If you are a Stanford PhD student the machinery sets you up pretty well.
    Feb 14 3
    • Salesforce tensorchic
      This too.
      Feb 14
    • Google BeataMaria
      You don’t need to be a PhD student, Elizabeth Holmes smashed that glass ceiling
      Feb 14
    • Yahoo k_
      This
      Feb 14
  • Salesforce tensorchic
    Have to follow this one.
    I've talked to and heard him few times, and he's got the risk taking mentality, out of the box thinking and certainly doesn't fit the mould, which lots of academics tend to do.
    Feb 14 9
    • Google BeataMaria
      They say these things about every up and coming corporate exec you know...
      Feb 14
    • Salesforce tensorchic
      Exactly. That's why those people become corporate execs. Not to say that execs are admirable or everyone seeks to become an exec.
      Feb 14
    • Google / Eng gangmember
      @tensorchic, what kind of rockstar research has he done?
      Feb 14
    • Salesforce tensorchic
      I didn't say he did. I also didn't say that he's a research rockstar. He's good at outreach, educating, seeing vision for products and companies and direction of the field and advising people. That combined with thorough knowledge of the field and Stanford and UCB network (most importantly!) Is more than enough to be "head of AI" Somewhere. They aren't looking for a michael Jordan or Terrence tao kind of person you know.
      Feb 14
    • Google / Eng gangmember
      Sorry, I completely misread your original response. Agree with what you've said.
      Feb 14
  • Google / Eng gangmember
    All salesmanship, very little actual skill. He'd be a decent T5 research scientist at Google or Facebook, but nothing more than that.

    He did not help in Deep Learning resurgence and only jumped on the bandwagon a year or two after it took off. During his Ph.D. he published about a couple of decent papers on automated captioning systems, but I don't recall them being anything mindblowing. And that was it... He went to OpenAI, didn't do shit there, and then joined Tesla. How? Because throughout his mediocre career he kept making blog posts and open-sourcing the Stanford course that he co-taught, all of which became very popular.

    He did amazing work teaching Deep Learning, but there's very little technical skill about him.
    Feb 14 0
  • New / Eng SharmaG
    What do people get out of bashing others? Karpathy is extremely influencial in his research: https://scholar.google.com/citations?user=l8WuQJgAAAAJ&hl=en

    Most profs can't hold a candle to his research influence over the course of their career, let alone matching it in 6 years.
    Feb 14 16
    • New / Eng SharmaG
      If that is the foundation of your argument, I understand it. Yes, you don't want to put up anyone on a pedestal that they could use to promote self interests. But scientific advancement requires a combination of trust and scepticism. That is the reason we have peer reviewed publications and citations as the basis of scientific advancement. Karpathy clearly won in those metrics and I respect him for it even if I don't buy everything he sells tomorrow.

      Asides, I do have qualitative arguments in favor of his work but I chose not to pursue that direction because you don't need them when you have data and numbers.
      Feb 16
    • Google rnmnrnmn
      How do you account for the quality of publications? Won’t there be a bias towards quantity over quality that can be easily abused? And how do you ensure the objectivity of the peer reviewers? I read the climategate emails and they were completely politicized.

      It seems the great scientific advances of the past did not rely on such metrics.
      Feb 16
    • New / Eng SharmaG
      This goes to the heart of democratic principles. Every person comes with their own individual views on every topic of interest. How do you know which opinion is right? Democracy solves that by going with the opinion that gets the most votes. Granted that it can be broken and gamed at times, but by and large it works. This is how political, scientific and economic fields are setup.

      In academia, citations are valued a lot for that reason. It used to be that the number of publications was a barometer of ones contributions. The community has gravitated to citation count and the h-index now. While they may not be perfect, it works. Citation in particular means that many people tried out someone's ideas and saw that it works or kinda works or agree that it is important. That kind of scrutiny by fellow community members is a very important validation based on democratic principles. There are always detractors to every one. Even MLK, Gandhi and Mother Theresa have detractors. But if a lot of people accept that someone is good, it's likely that there is something important about that person.
      Feb 16
    • Google rnmnrnmn
      Would good and evil exist in a world without humans?
      Feb 16
    • Google / Eng gangmember
      @rnmnrnmn: nah, he's not trolling, just completely oblivious about machine learning or research in general.
      Feb 18
  • New / Eng Bulldoze
    By being good with Twitter.
    Feb 14 0
  • Uber ug
    Salesmanship
    Feb 14 2
    • Salesforce tensorchic
      Exactly. Combined with theoretical knowledge, is pretty rare.
      Feb 14
    • Oracle blahbl
      OP
      How did he sell or advertise himself?
      Feb 14
  • Walmart.com salamanca
    Why does Tesla need AI? And did something useful come out of Tesla's AI org?
    Feb 14 3
    • Oracle blahbl
      OP
      Self driving cars?
      Feb 14
    • Walmart.com salamanca
      Auto pilot =/= self driving.
      Feb 17
    • Oracle blahbl
      OP
      Okay, cars don’t autopilot themselves. Someone has to 🤦🏻‍♂️
      Feb 17
  • Ahh good ole glorified matched filtering with a half space and a smearing of shit and heuristics.

    It is almost never about what you can do, it's all about self promotion. Always Be Closing. This holds for all industries and fields, the only ones making real contributions are the 6 sigmas.
    Feb 15 0
  • Google hfHs36
    Head of AI sounds like a PR thing
    Feb 14 0