9
Ranchu
32d

Did the MIT seriously just take down one of their own huge tiny image datasets for machine learning because of the same problem as GitHub's Master branch?

Comments
  • 1
    Do you has link?
  • 1
  • 4
    https://venturebeat.com/2020/07/...

    Apperently the accompanying descriptions was racist.
  • 4
    @Voxera
    I could definitely see how those samples might lead to subjective network biasing with the wrong composition.
  • 5
    I guess they needed a press story before they release their ten-billionth crappy peltier dehumidifier and claim that THIS one will somehow give third world countries water by defying the laws of thermodynamics.

    Seriously, seems like that's all MIT do these days
  • 1
    @PlatinumFire
    Well, they're a technology school whose Nobel prizes and staff laureates are primarily in economics. So take that for what it's worth.
  • 4
    They got the images by scraping search engines, and weren't expecting offensive content?
  • 5
    "The data set also contained pornographic content like non-consensual photos taken up women’s skirts" - This part i fully understand shouldnt be used at all.

    "such as nearly 2,000 images labeled with the N-word, and labels like “rape suspect” and “child molester.”

    Kinda understandable why they would take it down, these kind of people that would be offeneded over these kinds of things, would probably be burning down MITs buildings, assulting their employees and much worst stuff.

    Only stupid people get offended by things like this.
  • 1
    @Frederick
    Wouldn't even be the first time it happened
    https://wired.com/story/...
  • 0
    @Frederick Not only stupid people is offended.

    But a bigger problem is idiots that would take such a list from a reputable organization as an excuse to continue to discriminate not to mention that the display of such images and captions is illegal in many cases and would open them up to legal liabilities.
  • 0
    @Voxera Well as i said, i think its a fair reason to take it down due to how some of the photos in the data sets are straight up harassment of women and as far as i understood many of the images in the difference datasets has been collected illegally.

    But getting mad over some stupid lebels names is ridiculous and waste of time. Even though the naming is pretty childish unprofessional.

    Solutions: Rename the labels themselves or dont use the labels at all? problem solved.
  • 1
    @SortOfTested MIT also have good engineers, but they end up dead real quick
  • 4
    @cafebabe
    MIT engineers tend to be the brand of autist that has no issue with focus, but a deficit of interest. They're generally unconcerned with the human side of their engineering and just want to get the idea out of their head and into the world, regardless of how it's used. Given that, I agree with your statement.
  • 1
    @SortOfTested Sorry for that really bad joke about A. Swartz, which I respect very much.

    I agree with your take on it
Add Comment