Let’s assume I post a screenshot of a ChatGPT session on social media, and OpenAI sues me for the content.
That hypothetical doesn't have much to do with this indemnification clause. OpenAI wouldn't be the one filing a lawsuit against you. They are the ones being sued by someone else who saw the screenshot you posted.
OpenAI would just send you the bill once the case has been settled (because according to the ToS you agreed to defend them from lawsuits related to your use of ChatGPT).
Don’t they have to prove first that it actually is such a screenshot, and not a fake?
Yes, and during the whole process the prosecutor will force OpenAI to search through their logs/databases and turn over any evidence related to the case. It probably wouldn't take long since the screenshot would probably include the prompt from the User and they would just have to search for that.
Somehow this rings strangely similar to copyright cases against OpenAI, now with reversed roles. Who owns the authorship, how can we tell?
So far the courts have ruled that AI can't claim copyright to anything. The "prompter" could claim the copyright but they would also have to alter the output in some way to make it their own (at least as far as AI art is concerned, I assume it would be similar for copyright on text).
Sometimes, I wish Lemmy had a feature where I could quickly edit my comment within the first minute without receiving an "edit" flag... But even then my MS DOS machine wouldn't be able to edit it fast enough anyway.
I hate it when people say that I'm "throwing my vote away by voting for a 3rd party". If everyone voted for the person they actually liked, rather than the person who's likely to beat the other large party, maybe we'd see some better choices.
Hypothetical on this one, if the reason they decided to look into this was because they saw someone's post on social media about ChatGPT being able to reproduce parts of some copyrighted work, ChatGPT could bill the user for publishing that info.
It doesn't even have to be the sole reason for them to look into it. Technically they could bill anyone who posted content if that content wound up being used as evidence against OpenAI in any way (as I understand it, that's where the "relating to your use of the Services" part could be used).
But if I have misunderstood something about this hypothetical, please feel free to correct me.
I agree, it seems ridiculous, but according to the attorney in the video this would be enforceable, at least in the U.S.:
https://piped.video/fOTuIhOWFXU?t=330
I'm sure you could try to get your own attorney to try to fight back against OpenAI's attempt to bill you, but that's going to cost you as well.
Well, at least you have additional protections in your country.
For those in the U.S. this document is binding enough (at least according to this lawyer: https://youtu.be/fOTuIhOWFXU?t=330).
Edit: As with anything I'm sure you could argue in court that you shouldn't be held responsible for their legal bills and hopefully you would win, but that would still require you to go to court over the matter.
Right, but what about the case where you post something that doesn't contain lies at all?
What if ChatGPT outputs something that a certain former president gets offended by and he decides to sue OpenAI?
According to their ToS it doesn't matter if it's a "frivolous lawsuit". If OpenAI had to pay any attorney fees just to respond to some ridiculous lawsuit, they could still bill you for those costs.
I don't think it makes sense at that point at all.
Of course the vast majority of users would never have to worry about this, but it's still something to be aware of.
Basically just be careful if you like to post images/text taken straight from ChatGPT.
If you post anything that someone gets offended about and decides to sue ChatGPT (OpenAI) over it, they can turn around and bill you for those legal costs (whether they win the lawsuit or not).
Or if you post a screenshot that proves that you can get ChatGPT to write out the entire first chapter of some copyright protected book...
I've also seen people who like to "jailbreak" ChatGPT and then post things like tricking ChatGPT into giving instructions on how to make certain illegal devices and such. Again, just be careful and think if someone could sue the makers of ChatGPT and they include your social media post in the lawsuit, you have already agreed to pay their legal costs for that lawsuit.
The crypto example was only a suggestion because they have simply solved the exact same problem we are looking at: duplicate votes (transactions) and verifying the results while being able to hide it.
I would love to hear any other suggestions that people may have that solve these problems. Copying open source code from crypto isn't the only option. So let's look for solutions instead of dismissals (unless you're arguing for keeping votes public of course).
Let me be a little more clear, the Admins of your account's particular instance should be the only ones that have access to your votes.
Now the question remains about when your account posts/comments into a different instance, who should have access to those votes? Perhaps your instance has a way of obfuscating the votes of any user coming from your instance, or else only the admins of the community that you're posting into will have access to your votes?
The problem really comes down to how we avoid the problem with duplicating votes. Currently this is easy as each vote is public so every instance can verify the correct vote count. But implementing either of the solutions above will need a way to verify the correct number of votes.
To top it off you would also need a way to detect if a malicious instance had come along and started lying about how many votes had been cast.
One thing we can look at under the hood would be how cryptocurrency works as they have solved both the problem of duplicate values as well as the ability to trust those values being sent. All of the code is free and open source so we can pick out the parts that we need and reuse it. (And no, I'm not telling people to go out and buy crypto).
Z Cash would be a particularly good one to look at as it ensures a "zero knowledge" (or "zero trust") method of sending the values across "nodes" (or in our case "instances"). Using this, who is voting on what would be hidden, but we could ensure that the values are correct.
Additionally you could probably throw out the second hashing algorithm altogether and just keep the Blake2b hashing algorithm as this one is far more efficient and quick to compute (and that second algorithm was mostly thrown in to prevent people with specialized hardware from being able to come in and beat anyone else running on just a GPU/CPU). https://github.com/zcash/zcash
However, using this particular method would make it so that not even the instance admins would be able to view the details of anyone's votes (which may be a good thing after all if we decide that any random instance admin is not to be trusted).
Ideally, detecting bots should be up to the Admins. They should have access to the vote information, and they can share the tools with other admins to detect it. But the average user should not have unrestricted access to this data.
I'll have to find it, but I remember seeing some comments/ posts explaining how it was Huffman himself who was really the one pushing to kill off 3rd party apps.
Plus, he's the one who was lying about the Apollo dev's comments as well as editing other users comments.
QuadratureSurfer @ QuadratureSurfer @lemmy.world Posts 27Comments 494Joined 2 yr. ago
I'm guessing you're not a 90s kid...