Confounding the skimmers

Discussion in 'Ham Radio Discussions' started by K0HB, Sep 7, 2019.

ad: L-HROutlet
ad: l-rl
ad: Left-3
ad: abrind-2
ad: Left-2
ad: L-MFJ
ad: Subscribe
  1. KA8NCR

    KA8NCR Ham Member QRZ Page

    The astonishment is only matched by the hubris of Uber testing of self-driving cars on public roads.
  2. WF4W

    WF4W Ham Member QRZ Page

    it is odd that we used to be so conscious of our privacy and now, in the quest for likes and brags, we freely and wontonly give up our locations, friends, politics, etc. to anyone and eveyone that wants to see it...
  3. W9BRD

    W9BRD Ham Member QRZ Page

    I think self-driving cars should be painted fluorescent orange by law.
    K1LKP and KP4SX like this.
  4. WN1MB

    WN1MB XML Subscriber QRZ Page

    ...right before being placed in the crusher.
  5. KP4SX

    KP4SX Premium Subscriber QRZ Page

    It might come to that. In spite of all the great technology I think this loaf should remain in the oven for a while longer.
    K2CAJ and K1LKP like this.
  6. W3WN

    W3WN Premium Subscriber QRZ Page

    Great technology? Nah. Proof that just because something CAN be done, doesn’t mean it SHOULD be done.

    Trust me on this... if you see one coming, and about to turn into the crosswalk that you are walking through, never, EVER, drop something and then bend down to pick it up. Fortunately, I saw the fool thing turning just in time to jump back out of the way, as I have no desire to be on the 11 PM news.
    N2EY and KP4SX like this.
  7. W9BRD

    W9BRD Ham Member QRZ Page

    The concept of self-driving cars is always sold in terms of what we don't like about human driving, which is the 0th stupidity in the drumbeat for their adoption. It turns out that what a human brings to the intelligence of the driving process is immense perceptive, computational, learning, and decision-making power, beginning with the fact that life instinctively protects life, including its life and other lives. The smartest computer is still a psychopath; it cares about nothing and cannot care about anything.
    K2CAJ likes this.
  8. N2EY

    N2EY Premium Subscriber QRZ Page

    The smartest people will tell you there is a limit to what intelligence can do. Stupidity, OTOH, has no such limit.

    Have you ever driven the Schuylkill Expressway at rush hour?

  9. W3WN

    W3WN Premium Subscriber QRZ Page


    Many of those drivers aren’t human. They’re something else, all right... :p
    N2EY likes this.
  10. K2CAJ

    K2CAJ XML Subscriber QRZ Page

    On top of that, computer programming entails the careful eradication of weird corner cases. You must guarantee that your program provides the correct output for all valid inputs, and that it can also handle any possible input without behaving terribly.

    A lot of security vulnerabilities and attacks are possible because someone didn't account for a weird input, for example the Heartbleed bug { please send back the first 50,000-characters of "HELLO" }. Attackers can create those conditions and cause a disaster---or sometimes corner cases can just happen, like Telsas accelerating into firetrucks, or a malfunctioning stall sensor in the Boeing 737 Max.

    But what happens when we replace a precisely defined program with AI? Now you input a bunch of camera and sensor inputs, and the trained AI recognizes an obstacle. What are the corner cases? What weird inputs are not being accounted for in the training data, and will cause a false decision? What is the process for guaranteeing that the program works in general, and not just a common case?

Share This Page