Richard Kelley

@richardkelley

Roboticist. Using machines to understand intelligence. Occasionally making cars drive themselves.

Reno-Tahoe
Vrijeme pridruživanja: travanj 2008.

Tweetovi

Blokirali ste korisnika/cu @richardkelley

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

  1. proslijedio/la je Tweet
    2. velj

    Has any human driver, in the history of driving, actually encountered a "trolley problem" kind of situation? I would be interested in hearing some actual examples.

    Poništi
  2. 15. ruj 2019.

    The joys of robotics: Spent yesterday hunting for bugs in a new model predictive controller for the MKZ, only to discover that the code works but the hardware starts to fault when the trunk gets hot. Time for more AC I guess...

    Poništi
  3. 18. kol 2019.

    I’m just getting started, but it seems like Julia could be a great language for doing robotics. I’m looking at it as a potential Python replacement.

    Poništi
  4. 16. kol 2019.

    This is a great list of questions to ask about AV demos/“deployments.” And for the most part you can replace “AV” with “robot” and the questions remain extremely useful.

    Poništi
  5. proslijedio/la je Tweet
    14. kol 2019.
    Poništi
  6. proslijedio/la je Tweet
    7. kol 2019.

    While "minor" the AV backed into a car. Driver "had taken control" -- but maybe couldn't compensate in time? They should publish an analysis to build/maintain trust. A claim that low speed means "never" for fatalities seems optimistic.

    Poništi
  7. 31. srp 2019.

    I think stuff like this is a great argument for combining detections from multiple sensor types - lidar+cameras for example. I have yet to see an attack that would reliably work against a multimodal detection pipeline.

    Prikaži ovu nit
    Poništi
  8. 31. srp 2019.

    Fun article about attacking mobileye object detection using a drone. I actually have the camera from the paper. I stopped using it in my AV because it's unreliable - usually very good but occasionally catastrophically bad.

    Prikaži ovu nit
    Poništi
  9. proslijedio/la je Tweet
    10. lip 2019.
    Poništi
  10. proslijedio/la je Tweet
    29. svi 2019.

    Team CERBERUS from shared video of the system of systems solution it tested during STIX. The team's walking & flying robots collaborate to facilitate exploration in sensing-degraded & communications-constrained subterranean environments.

    Poništi
  11. 22. svi 2019.

    NASA is doing some very cool drone testing in downtown Reno:

    Poništi
  12. proslijedio/la je Tweet

    Given infinite data, you can solve arbitrarily complex problems with a "dumb" model (such as a densely-connected neural network). However, when it comes to difficult real-world problems like self-driving, "lots of data" is very different from "infinite data". The long tail is fat

    Prikaži ovu nit
    Poništi
  13. proslijedio/la je Tweet
    14. svi 2019.

    A worrying example of the fallacy of big data. There are 'countably infinite things' that can possibly happen and just 500,000 cars ... around 25% of which are driving around California. We need a lot more than just big data to have safe autonomous vehicles.

    Poništi
  14. 24. tra 2019.

    I'm excited to be working on a project to explore how blockchain tech can be used to help secure autonomous systems and smart infrastructure:

    Poništi
  15. proslijedio/la je Tweet
    6. tra 2019.

    Our robots went in, explored and came back safe during the Challenge - Video from the Aerial Scouts of - with HiPeR-UCBerkeley

    Poništi
  16. 18. ožu 2019.

    Sutton's essay is interesting, but his argument is a lot less convincing if your definition of AI includes robots with complex dynamics in the real world, where human-designed algorithms still vastly outperform learning

    Poništi
  17. 9. ožu 2019.
    Poništi
  18. proslijedio/la je Tweet
    8. ožu 2019.

    Just put up a new post, where I explain the intuitions behind Natural Gradient learning: how to connect high-level concepts with the often-difficult-to-parse math. Herein lie gory details of optimization algorithms, and also puns!

    Poništi
  19. proslijedio/la je Tweet
    28. velj 2019.

    When people say 'we use in the business process.' Sometimes...I know that...🙄

    Poništi
  20. 28. velj 2019.

    Parking the autonomous car at Nevada's legislature while I talk about AI for transportation in the state.

    Poništi

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:

    ·