New research shows driverless car software is significantly more accurate with adults and light skinned people than children and dark-skinned people.

  • MyTurtleSwimsUpsideDown@kbin.social
    link
    fedilink
    arrow-up
    4
    arrow-down
    5
    ·
    edit-2
    1 year ago

    Wow. that’s all kinds of incorrect

    It’s not a discriminatory bias or even one that can really have anything done about it.

    It is absolutely data training bias. Whether it is the data that ML was trained on or the data that programmers were trained on is irrelevant. This is a problem of the computer not recognizing that a human is a human

    It’s purely physics.

    It is not. See below:

    Is it harder to track smaller objects or larger ones?

    No, not if the scale of your hardware is correct. A 3’ tall human may be smaller than a 6’ one, but it is larger than a 10” traffic light lens or a 30” stop sign. The systems do not have quite as much trouble recognizing those smaller objects. This is a problem of the computer not recognizing that the human is a human.

    Is it harder for an optical system to track something darker. In any natural scene.

    Also no. If that were the case, then we would have problems with collision bias against darker vehicles, or not being able to recognize the black asphalt of the road. Optical systems do not rely on the absolute signal strength of an object. they rely on contrast. A darker skin tone would only have low contrast against a background with a similar shade, and that doesn’t even account for clothing which usually covers most of a persons body. Again, this is a problem of the computer not recognizing that the human is a human.

    smaller and darker individuals have less signal. Less signal means lower probability of detection,

    No, they have different signals. that signal needs to be compared to the background to determine whether it exists and where it is, and then compared to the dataset to determine what it is. This is still a problem of the computer not recognizing that the human is a human.

    It’s the same reason a stealth bomber is harder to track than a passenger plane. Less signal.

    Close, but not quite.

    1. In this case the “less signal” only works because it is compared to a low signal background, creating a low contrast image. It is more like camouflage than invisibility. Radar uses a single source of “illumination“ against a mostly empty backdrop so the background is “dark”, like looking up at the night sky with a flashlight.
    2. The less signal is not because the plane is optically dark. It has a special coating that absorbs some of the radar illumination and a special shape that scatters some of the radar illumination, coming from that single source, away from the single point sensor. Humans of any skin tone are not specially designed to absorb and scatter optical light from any particular type of light source away from any particular sensor. Even at night, a vehicle should have a minimum of 2 headlights as sources of optical illumination (as well as streetlights, other vehicles. buildings, signs and other light pollution) and multiple sensors. Furthermore, the system should be designed to demand manual control as it approaches insufficient illumination to operate.

    This is a problem of the computer not recognizing that the human is a human.