3 ms·

Personally, I feel it would be a much smaller problem if Tumblr had an internal AI thing going on. What users REALLY don't like is that they have confided a pos

Personally, I feel it would be a much smaller problem if Tumblr had an internal AI thing going on. What users REALLY don't like is that they have confided a post with one website, and that website just shared that post with a third-party because if opens up infinite possibilities. If Tumblr can take your post and give it to OpenAI, they can take your post and give it to anyone, and that's the problem. Because for users, what they post is "between them and Tumblr" and not anybody else. I'll even say more. Artists don't care if you scrape their art to make AI generated art with it. Because when this happens, it's "between artists and scrapers" and not anybody else, so it's fine. What they do care about is when people post that AI generated art on the internet, or publish it professionally, or do basically anything with it. In other words, there's a sense of privacy when there is only two parties involved, no matter what is going on, but the instant a third party gets into the system, the first party will freak out, because then you lose that sense of privacy.

nonrandomstring 3y ago
You're saying consent isn't transitive. That's why they always ask for (or more frequently just take) "a paid up, non-exclusive, irrevocable, worldwide license"
- AlienRobot 3y ago
  I think "consent" is not the right word for this, because it's never simply "consent" it's always boundaries and expectations. Consent implies there's always one yes/no answer for a process. In practice the process is always so complicated that it would require countless yes/no questions, to the point nobody sane would want to deal with it. Just look at cookie banners and imagine if we had a consent popup for every single thing that needs to be download to show a page: do you consent to download the HTML? Do you consent to download the CSS? Do you consent to download the Javascript? Do you consent to download the images? The user just wants the see the page. You have to make assumptions about how far that consent goes, so it's absolutely transitive, the problem is at what point it crosses the line.
  - nonrandomstring 3y ago
    So, please forgive the necessary parallel here. Let's think about dating and imagine if we both had to consent to every single thing that needs to happen to get it on. Do you consent to dinner with me? Do you consent to coffee at yours? Do you consent to us getting our clothes off? Do you consent to ... you get the picture. Thing is, we actually do this, but silently. The "assumptions" you rightly speak of are the mutual trust mechanisms that allow two adults to make clear shortcuts without explicit words. We move at a certain speed. We check for feedback. We use clever signals. We base our trust model on the NSA definition that "Trust is the ability to do harm" in context where there is some symmetry. Sex is actually risky for both parties. And no means no. But in the digital world we're talking about a massive power asymmetry. When you're with a big corporation it moves from dinner to data-rape in a few milliseconds. No means "yeah but harder". Consent doesn't disappear because the process got complex. Human relations are more complex than any web page. It's about whether we obey the rules and show mutual respect. Big-tech corporations absolutely do neither. and that's the problem here, not complexity. What might level the game, is if when I visit your site with my web browser, you risk that I could do you some serious harm.
    - AlienRobot 3y ago
      Unfortunately, it's not a human on the other side, it's a system, whether it's a system built based on business rules or a system built based on computer algorithms, everyone gets the same inflexible system. It's not logistically feasible to create a custom experience for each user based on how far they consent to every little thing, as the use cases are infinite: you may consent to download CSS on this site, but not on that site, or on that site, but not on pages that match this regex, etc. If you tried to do that, your product would immediately flop because most users would find it too annoying and complicated to use compared to any system that makes reasonable assumptions about consent. I agree that many modern software have no respect to humans, but that's not because the developers didn't want to implement a yes/no consent dialog or several of them, it's because they think the only boundary they can't cross is what the law says they can't, and everything else is fair game.
      - 3y ago

nonrandomstring 3y ago

You're saying consent isn't transitive. That's why they always ask for (or more frequently just take) "a paid up, non-exclusive, irrevocable, worldwide license"

AlienRobot 3y ago
I think "consent" is not the right word for this, because it's never simply "consent" it's always boundaries and expectations. Consent implies there's always one yes/no answer for a process. In practice the process is always so complicated that it would require countless yes/no questions, to the point nobody sane would want to deal with it. Just look at cookie banners and imagine if we had a consent popup for every single thing that needs to be download to show a page: do you consent to download the HTML? Do you consent to download the CSS? Do you consent to download the Javascript? Do you consent to download the images? The user just wants the see the page. You have to make assumptions about how far that consent goes, so it's absolutely transitive, the problem is at what point it crosses the line.
- nonrandomstring 3y ago
  So, please forgive the necessary parallel here. Let's think about dating and imagine if we both had to consent to every single thing that needs to happen to get it on. Do you consent to dinner with me? Do you consent to coffee at yours? Do you consent to us getting our clothes off? Do you consent to ... you get the picture. Thing is, we actually do this, but silently. The "assumptions" you rightly speak of are the mutual trust mechanisms that allow two adults to make clear shortcuts without explicit words. We move at a certain speed. We check for feedback. We use clever signals. We base our trust model on the NSA definition that "Trust is the ability to do harm" in context where there is some symmetry. Sex is actually risky for both parties. And no means no. But in the digital world we're talking about a massive power asymmetry. When you're with a big corporation it moves from dinner to data-rape in a few milliseconds. No means "yeah but harder". Consent doesn't disappear because the process got complex. Human relations are more complex than any web page. It's about whether we obey the rules and show mutual respect. Big-tech corporations absolutely do neither. and that's the problem here, not complexity. What might level the game, is if when I visit your site with my web browser, you risk that I could do you some serious harm.
  - AlienRobot 3y ago
    Unfortunately, it's not a human on the other side, it's a system, whether it's a system built based on business rules or a system built based on computer algorithms, everyone gets the same inflexible system. It's not logistically feasible to create a custom experience for each user based on how far they consent to every little thing, as the use cases are infinite: you may consent to download CSS on this site, but not on that site, or on that site, but not on pages that match this regex, etc. If you tried to do that, your product would immediately flop because most users would find it too annoying and complicated to use compared to any system that makes reasonable assumptions about consent. I agree that many modern software have no respect to humans, but that's not because the developers didn't want to implement a yes/no consent dialog or several of them, it's because they think the only boundary they can't cross is what the law says they can't, and everything else is fair game.
    - 3y ago