• Posts
  • RSS
  • ◂◂RSS
  • Contact

  • Conversation with an AI Researcher

    July 20th, 2017
    airisk, giving  [html]
    Earlier this week I had a conversation with an AI researcher [1] at one of the main industry labs as part of my project of assessing superintelligence risk. Here's what I got from them:

    They see progress in ML as almost entirely constrained by hardware and data, to the point that if today's hardware and data had existed in the mid 1950s researchers would have gotten to approximately our current state within ten to twenty years. They gave the example of backprop: we saw how to train multi-layer neural nets decades before we had the computing power to actually train these nets to do useful things.

    Similarly, people talk about AlphaGo as a big jump, where Go went from being "ten years away" to "done" within a couple years, but they said it wasn't like that. If Go work had stayed in academia, with academia-level budgets and resources, it probably would have taken nearly that long. What changed was a company seeing promising results, realizing what could be done, and putting way more engineers and hardware on the project than anyone had previously done. AlphaGo couldn't have happened earlier because the hardware wasn't there yet, and was only able to be brought forward by massive application of resources.

    In their view, it doesn't make sense to try to influence where the field will go more than a few years out. If an area has been underinvested in by people focusing elsewhere, then after a few years, with faster hardware, that area will have lots of (now) low hanging fruit and quickly catch up. This would imply that a strategy of differential technology development, where you're trying to change the relative rates at which parts of AI advance by working in a part you think is likely to make us safer, wouldn't work very well.

    (It looks to me like a big difference between (my model of) their view and (my model of) Dario's is, what fraction of the best research directions get pursued? The lower you think that fraction is then the less the the "underinvested stuff will catch up when hardware gets better" view fits. This is also another connection back to technological distance: if you think underinvested stuff naturally starts to look more promising and catch up as people go into it, then the farther we are from AGI in terms of remaining work then the less a differential technology approach helps.)


    [1] They're a friend of mine through non-EA connections, so this is more like drawing a sample from the pool researchers as a whole. At their request, I'm not using their name, affiliation, or gender.

    Comment via: google plus, facebook

    Recent posts on blogs I like:

    What should we do about network-effect monopolies?

    Many large companies today are software monopolies that give their product away for free to get monopoly status, then do horrible things. Can we do anything about this?

    via benkuhn.net July 5, 2020

    More on the Deutschlandtakt

    The Deutschlandtakt plans are out now. They cover investment through 2040, but even beforehand, there’s a plan for something like a national integrated timetable by 2030, with trains connecting the major cities every 30 minutes rather than hourly. But the…

    via Pedestrian Observations July 1, 2020

    How do cars fare in crash tests they're not specifically optimized for?

    Any time you have a benchmark that gets taken seriously, some people will start gaming the benchmark. Some famous examples in computing are the CPU benchmark specfp and video game benchmarks. With specfp, Sun managed to increase its score on 179.art (a su…

    via Posts on Dan Luu June 30, 2020

    more     (via openring)


  • Posts
  • RSS
  • ◂◂RSS
  • Contact