5 ms·
It seems exceptionally unlikely to me that OpenAI would be "reading user prompts". More likely is a leak somewhere else. Obviously Anthropic/OpenAI are engaged
by onidj 24d ago
It seems exceptionally unlikely to me that OpenAI would be "reading user prompts". More likely is a leak somewhere else. Obviously Anthropic/OpenAI are engaged in espionage stuff with each other. I am guessing Levent just mentioned something to someone at Anthropic and it got out.
- simianwords 24d agowhy is this downvoted? do people really think OpenAI is snooping at people specifically? like they are looking for good leads into new problems or ideas and they found this guy's codex thread and used it? come on man, even for conspiracy theories this is stupid.
- gcr 24d agothis is a highly marketable problem and specific teams at OpenAI were aware of specific competitor efforts. I would be surprised if this were happening on a large scale, but 1. Less than a hundred people in the world are working at this problem, 2. A significant fraction of those happen to work at competing hyperscalers, 3. Those hyperscalers repeatedly show themselves not to take user privacy seriously
- simianwords 24d agoI'm not sure what your points 1 and 2 have to do with anything. Both directionally increase the probability of hyperscalers also finding the solution independently. > Those hyperscalers repeatedly show themselves not to take user privacy seriously where? Any examples?
- hn_acc1 24d agoHave you been living under a rock? None of the big tech companies care one bit about user privacy in the US.
- CamperBob2 24d agoThat's not an answer.
- hellohello2 24d agoAny degree of tracking what people do is unprivate. Every single web interaction you perform is tracked. All LLM companies store all your conversations by default. Do you need more examples?
- freejazz 24d agoIt's one of those questions that sets the bar so low, the conversation might as well not be taking place
- HDThoreaun 24d agoYes that sounds like something openAI would do to me. Not that they’re just looking through random professors chats but they heard buckmaster made progress on navier stokes and decided to read his chats.
- ajnin 24d ago> https://openai.com/policies/how-your-data-is-used-to-improve-model-performance/ https://openai.com/policies/how-your-data-is-used-to-improve... They explicitly say that they use your "content" to improve their models. Considering they practically have infinite compute at their disposal, why is it surprising that they would look for juicy data in there to make them look good ? When they ingested basically the entirety of human knowledge without regard to the rights of others, when they burn books by the thousands, when their relentless barrage of bots have rendered the Web borderline unusable, why would they stop at that line ?
- simianwords 24d ago[flagged]
- GPerson 24d agoYou’re always going around insulting people. Does it make you feel powerful? Is it really hard to believe the people who would consume all the world’s data regardless of copyright and norms and permissions would not respect the data privacy of a user?
- Panoramix 24d agoYes, that would be par course with OpenAI behavior
- hn_acc1 24d agoEven if they claim not to use it, they're probably using it and hoping they don't get caught. They have zero ethics or morals, they just want to "win" to get mega-rich.
- jawilson2 24d ago> do people really think OpenAI is snooping at people specifically yes. Rules and laws are for the poor.
- dgellow 24d agoObviously, yes
- freejazz 24d ago> do people really think OpenAI is snooping at people specifically? I'd be shocked if they weren't
- rockdoe 24d agoOf course they are "reading user prompts" in the sense that it's being used to further training. That's why you have to pay up to opt out of that. That's also why they refused to answer that question: the answer is obviously "obviously"!
- lukewarm707 24d agoif training is on, they are reading user prompts. training is on by default.