7 ms·

If your data is stored in a database that a company can freely read and access (i.e. not end-to-end encrypted), the company will eventually update their ToS so

• sandworm101 3y ago
  Then it is time to take control by seeing such databases with false information. We could probably design data specifically to attract an AI system looking for well-documented data. I wonder how images of Steven Colbert uploaded to imgur it would take to convince the AI that he actually was.
  • nwoli 3y ago
    I don’t think this is a possible approach. It would basically mean needing to feed in so much noise that a human wouldn’t be able discern reality from fiction given no priors. Modern ML is too smart
    • BiteCode_dev 3y ago
      AKA encryption.
    • proser 3y ago
      Given how easy it is to keep an average human from discerning reality (see the last decade of boomers/gen xers on Facebook as an example) and the massive potential to create slight variations with LLMs, I don't think a Stephenson-style misinformation propagation campaign is that far outside the bounds of possibility. That's a lot of compute power to waste on it, but I would guess that that's what bot networks are going to be used for in the future (or already are, right now, if they're done mining bitcoin).
    • sandworm101 3y ago
      That's why I said president of canada. A non-intelligent AI would not find any priors indicating who was the president. There isn't one. So a few thousand images in the database, against zero information to the contrary, might be enough.
• confoundcofound 3y ago
  Yep. The initial decision was certainly driven by many stakeholders who deeply believe this is a key advantage to acquire. Users are fighting an uphill battle, and no amount of pushback short of lost revenue will stop them. They'll find another way.
• karaterobot 3y ago
  Yeah, I think it's safe to assume this. One time as a contractor, I was working on a GIS-enabled game for some company. Their privacy policy was very clear that they will not use or sell your data for any reason. It was even on their home page, because it was a key selling point. By the time I was involved with this app, the money was running out, and I vividly remember being in a meeting with the leadership where they were enumerating their options, one of which was to find someone to buy all this user data they'd collected. Their commitment to privacy disappeared the instant it was tested. Ever since then, I just assume this is how every company works. No matter what they say, they're selling your data.
  • fhd2 3y ago
    That's a fair assumption in my experience from running companies. I've seen management with ethics, but even then, eventually new people are in charge, shares (or the whole thing) get sold... That's what contracts are for, commitments from a company are commitments from people that might not be around very long.
    • voakbasda 3y ago
      Most EULA contracts have a clause that expressly permits the company to abandon their obligations upon sale of the business or other such events. However, that seems somewhat redundant, since these same contracts also assert that they can modify the terms whenever they want. For the average user, there's little to no recourse when the original contract gets violated.
      • ronsor 3y ago
        Courts seriously need to crack down on these EULAs
  • crazygringo 3y ago
    This suggests that we require an explicit law that forbids changes in ToS from ever apply to data collected retroactively. With a real monetary punitive incentive that allows users to sue. I absolutely believe that companies should have the freedom to change their ToS moving forwards, and that "promises" to "never" change a ToS are worthless (your example is a perfect reason why). BUT, I simply don't see how it could ever be fair to use data retroactively. If you change your ToS, you should only get to monetize new user data going forwards. It seems like a basic legal principle. Is there any existing law/precedent that suggests this is already the case, i.e. that such a company can be successfully sued but that people don't usually try? Or do we need new laws around this, and are there any government reps pushing for this?
• damnesian 3y ago
  Our enterprise issued an ultimatum recently requiring no software outside of our MS contract. Which is going to be severely limiting, but painfully doable. This move really seems to be in reaction to the Zoom debacle. Now we'll be forced to use Teams for online trainings after pretty much universally using Zoom since the pandemic. Our customers are gonna love that. It's above my pay grade but I wonder if we've already signed our data over to MS for a certain price, with that stipulation.
• msla 3y ago
  Or they'll do it without updating the ToS. Nothing's stopping them. Adults realize other adults do what benefits them.
  • IshKebab 3y ago
    The GDPR is stopping them.
    • palata 3y ago
      But is it even enforced? Take software licenses. Most software out there uses open-source licensed dependencies. Even the permissive licenses require attribution. But most software does not honor those licenses, and nobody cares. Enforcing such a law takes money. I guess at least the GDPR can be enforced, to some extent, with Big Data. It seems like the fines are usually ridiculously low (they don't seem like an intensive for the company to change anything), but that's better than nothing.
      • IshKebab 3y ago
        It's not very well enforced but the fines are large, and big companies like Zoom will have suitably paranoid legal departments ensuring that it is followed.
• rsynnott 3y ago
  I mean, if they have a bunch of enterprise customers there’s a fairly strong disincentive; any sensible corporate customer, if they did this, would presumably say, right, cool, MS Teams time.
  • esperent 3y ago
    I'm having trouble parsing your comment. Are you saying MS Teams is more or less likely to use your data to train AI? I'm assuming it's safe because they make a big deal about compliance. But on the other hand MS have a huge incentive to obtain data for AI since they are going all in on it.
    • rsynnott 3y ago
      Microsoft is unlikely to _secretly_ do it, certainly; they would be sued into oblivion, and that’s before the EC gets to them (for GDPR enforcement grinds slow but exceeding fine, and also with exceedingly big fines). US regulators might also take an interest. Also, modern Microsoft’s _whole thing_, more or less, is “we are your trusted enterprise partner who definitely won’t do bad stuff with the data you put in our cloud services”. They are unlikely to throw that away for a bit of flavour-of-the-month AI boosterism. Note that they’ve recently released a private ChatGPT thing; they can’t credibly acknowledge the problem with one hand and exploit it with the other.
• simonw 3y ago
  I don't think that holds up. Paying customers absolutely HATE the idea of their data being used to train AI models without their permission - this Zoom story is just the latest example of that. Companies that try to do this will get burned. Zoom just got burned really badly, and I personally don't think they actually intended to even do this - they just didn't make it clear enough that they were NOT going to do it, which sparked a PR nightmare firestorm for them. I think the incentives for companies are very much the other way round: if paying customers hate this, then the incentives are NOT to do it.
  • Jeslijar 3y ago
    I promise not to use any of the data I collect about you! Except to improve services (ML training!), Advertising (We'll sell your data to advertisers!), and by government order (pick your favorite three letter agency.) Alternatively: They'll just use your data anyway and not tell you about it. How is anyone going to prove their data was used as the source? It's like all the NDAs people sign when they join a company and they pinky promise not to use it at the next job where they land a big fat raise and promotion... suuure they aren't going to take what they've learned and improve upon it to try and get more promotions and raises in the future. It can't be stopped.
    • josephg 3y ago
      > It can't be stopped. Maybe. Maybe not. Hard to tell without trying. I applaud the EU and California for giving it a go with their data protection laws. I really hope their crackdown on this stuff is effective.
    • sublinear 3y ago
      > It's like all the NDAs people sign when they join a company and they pinky promise not to use it at the next job where they land a big fat raise and promotion Uh no. An NDA would cover proprietary intellectual property, not tools everyone else also uses. Unless you're now working for the previous employer's competitor, it's unlikely that proprietary tech would ever be used. Working for competitors and partners is usually also forbidden for some period of time after leaving.
• blibble 3y ago
  no need to even update the ToS just call it "fair use", like OpenAI and GitHub
  • wredue 3y ago
    Yeah. And then regulars will argue for you that changing identifiers but otherwise copy and pasting code is perfectly cool fair use regardless of your codes licensing (I do not license any of my code for free corporate use, for example, but apparently, as long as chat GPT outputs the exact copy and paste of my code with a few things slightly changed, it’s cool).
  • neltnerb 3y ago
    I mean, around now would be a fairly convenient time for giant companies to change their mind about copyright. After all, that creative commons drawing and GPL code are only protected because of copyright, correct?
    • homarp 3y ago
      >a fairly convenient time for giant companies to change their mind about copyright but then they would not be able to do this https://news.ycombinator.com/item?id=37100140 https://news.ycombinator.com/item?id=37100140
      • neltnerb 3y ago
        I am speaking tongue in cheek of course, if they weren't themselves protected at all it would not work out for them. They couldn't own the distribution system well enough to keep the cat in the bag. But I can see them being okay with bringing copyright back to 10 years or something, they make most of the money from publishing a new television series early on. If they made it ten years, paying for streaming old television series doesn't require paying the copyright owners anything so it's pure profit for the streaming service. Of course people can go elsewhere, but if they can only get things made in the last 10 years on some particular service, they'd just use the same to view old things... and pay the website without the creators getting anything. So every distributor wins because they can all offer the older material as a pure profit for themselves and an enticement to join the service (on top of the actual unique material from the last 10 years).
• _Algernon_ 3y ago
  Legislation should step in to forbid applying new or changed ToS to data collected before that ToS was explicitly accepted by a user. Would be a PITA for the business (as they now have to track under what ToS version each piece of data is stored under), but hopefully enough of a PITA to make them treat that data as the liability it ought to be.
• seltzered_ 3y ago
  Sounds almost reminiscent of Jamie Zawinski's Law of Software Envelopment: "Every program attempts to expand until it can read mail. Those programs which cannot so expand are replaced by ones which can."
  • josephg 3y ago
    Huh? Barely any programs on my computer can read mail. That’s a silly law.
    • esperent 3y ago
      It's probably outdated. A modern version might be "every program expands until it becomes a platform with instant messaging capabilities and an app store..."
• stubish 3y ago
  I think most user data will be completely useless for AI training. No context, no quality control. Even assuming wisdom of crowds and on average you get the correct answer, you are training an AI to give a mediocre response. Back in the days of Big Data hype, many companies decided to store everything because data was valuable. But in most cases it wasn't worth the cost of the hard drives, because it only has value if there is a buyer willing to pay the cost of making use of it. There is a lot of data out there, but it is information and knowledge that has value.
• spacebanana7 3y ago
  Unlike consumers, many enterprises really care about data policies. Procurement, compliance, legal and audit teams inside enterprises often have the power to block purchases and contract renewals. Many of those departments actually read ToS too. This makes it difficult for B2B companies like Zoom to use customer data for AI training.