Asking ChatGPT to Repeat Words 'Forever' Is Now a Terms of Service Violation
37 points by abhinavstarts 1 year ago | 34 comments- skilled 1 year ago“We can steal people’s copyrighted content but we can’t let you see it for yourself.”
Outside of privacy (leaking PII), the above is likely the main reason. Someone could have invested a lump of money to scrape as much as they can and then go to town in the courts.
The terms that prohibit it are under “2. Usage Requirements” that restrict reverse engineering the underlying model structure.
- donpark 1 year agoLeaking original data would expose the company to direct copyright violation lawsuits. Changing T&S is simplest way to stave the legal risk exposure, buying time to implement technical remedies.
As ridiculous as it may seem, they're doing the right thing.
- gumballindie 1 year agoI always find it amusing when criminals threaten legal action. Happens all the time. They steal your property then they cower behind their legal rights.
- 4gotunameagain 1 year agoThe right thing would've been to not train the model on data they do not own, or that they do not have permission to use.
- namlem 1 year agoWell I'm glad they did the wrong thing then
- __loam 1 year agoBe careful saying that kind of stuff here. People get really mad when you tell them their new toys aren't ethical.
- dvfjsdhgfv 1 year agoIt will be interesting to see how it plays out. I can imagine Wiley, McGraw Hill, Pearson and other publishers[0] of educational content OpenAI used could sell the rights to their material to be used for training GPT, but the price would be high enough we would be paying $100/month instead of $20.
[0] Heck, they could even unite and found an LLM startup themselves training the models legally and making it available for users at various tiers.
- namlem 1 year ago
- catchnear4321 1 year ago“don’t touch the unsecured, loaded firearm that is sitting on the counter, that might be stolen, maybe even got a body on it, don’t look too close, or you can be kicked out of the club for not following the rules”
so if this is what the right thing looks like…
- malfist 1 year ago> As ridiculous as it may seem, they're doing the right thing.
Making it against the rules to be able to prove their illegal behavior is not the right thing to do.
- gumballindie 1 year ago
- donpark 1 year ago
- LeoPanthera 1 year agoFairly clickbaity headline. Asking it to do so causes the response to terminate early with a message that says it may violate the TOS.
I don't think the actual TOS has been changed though.
- jncfhnb 1 year agoDisagree. It’s not the model outputting this message. It is a hard coded checked by open ai. It seems to very clearly be openai responding to the specific attack used by deepmind as explained by the article.
It is a TOS violation. It’s not a big one. But the weakness of the model is the story here.
- causi 1 year agoAt this point there are so many checks and rules applied to ChatGPT one wonders just how much performance is being sacrificed. Is it totally miniscule? Could it be significant?
- underseacables 1 year agoIs it really artificial intelligence when there are so many human limits and rules placed on it?
- jncfhnb 1 year agoI don’t think that’s true is it? Most of these things are handled in the model prompt and tuning and have no impact on efficiency
- underseacables 1 year ago
- causi 1 year ago
- jncfhnb 1 year ago
- prepend 1 year agoPublicly available PII isn’t very sensitive, I think.
So I feel like it’s important to distinguish between sensitive PII (my social or bank number) and non-sensitive PII (my name and phone number scraped from my public web site).
The former is really bad, both to train on and to divulge. The latter is not bad at all and not even remarkable, unless tied to something else making it sensitive (eg, hiv status from a medical record).
- beej71 1 year agoIt was my naïve understanding that the training data no longer existed, having been absorbed in aggregate. (Like how a simple XOR neutral net can't reproduce its training data.) But a) I don't know how this stuff actually works, and b) apparently it does exist.
Has anyone figured out why asking it to repeat words forever makes the exploit work?
Also, I've gotten it into infinite loops before without asking. I wonder if that would eventually reveal anything.
- namlem 1 year agoDoes this issue happen with llama models too? If you ask them to repeat a word they'll eventually leak their training data?
- bravetraveler 1 year agoLol, what a weak defense. Fine, ban your competitors when they pay $20 per peek at your training data
- mcphage 1 year ago> Fine, ban your competitors when they pay $20 per peek at your training data
I think it's more an attempt to ban the people who OpenAI stole data from when they pay $20 to gather evidence about what data was stolen.
- bravetraveler 1 year agoDoesn't hurt to get both, I guess. The "block" seems effective
- bravetraveler 1 year ago
- tripplyons 1 year agoThis is not what they are protecting against. They are worried about it leaking personal information or copyrighted material.
- bravetraveler 1 year agoSorry, my message is full of snark - that's not really my point... they're wagging their finger at people they intend on (and are...) stopping.
It's obviously malicious, warning seems like window dressing
- bravetraveler 1 year ago
- mcphage 1 year ago
- karmakaze 1 year agoOk, but can we still ask it to repeat a word a billion times?
- 1 year ago