←back to thread

Fighting the New York Times' invasion of user privacy

(openai.com)
237 points meetpateltech | 2 comments | 12 Nov 25 14:08 UTC | HN request time: 0s | source
Show context
EdNutting ◴[12 Nov 25 15:03 UTC] No.45901100[source]▶
So why aren’t they offering for an independent auditor to come into OpenAI and inspect their data (without taking it outside of OpenAI’s systems)?

Probably because they have a lot to hide, a lot to lose, and no interest in fair play.

Theoretically, they could prove their tools aren’t being used to doing anything wrong but practically, we all know they can’t because they are actually in the wrong (in both the moral and, IMO though IANAL, the legal sense). They know it, we know it, the only problem is breaking the ridiculous walled garden that stops the courts from ‘knowing’ it.

replies(2): >>45901342 #>>45908047 #
glenstein ◴[12 Nov 25 15:26 UTC] No.45901342[source]▶
By the same token, why isn't NYT proposing something like that rather than the world's largest random sampling?

You don't have to think that OpenAI is good to think there's a legitimate issue over exposing data to a third party for discovery. One could see the Times discovering something in private conversations outside the scope of the case, but through their own interpretation of journalistic necessity, believe it's something they're obligated to publish.

Part of OpenAI holding up their side of the bargain on user data, to the extent they do, is that they don't roll over like a beaten dog to accommodate unconditional discovery requests.

replies(1): >>45901881 #
freejazz ◴[12 Nov 25 16:08 UTC] No.45901881[source]▶
>By the same token, why isn't NYT proposing something like that rather than the world's largest random sampling?

It's OpenAI's data, there is a protective order in the case and OpenAI already agreed to anonymize it all.

>Part of OpenAI holding up their side of the bargain on user data, to the extent they do, is that they don't roll over like a beaten dog to accommodate unconditional discovery requests.

lol... what?

replies(1): >>45903044 #
glenstein ◴[12 Nov 25 17:34 UTC] No.45903044[source]▶
Discovery isn't binary yes/no, it involves competing proposals regarding methods and scope for satisfying information requests. Sometimes requests are egregious or excessive, sometimes they are reasonable and subject to excessively zealous pushback.

Maybe you didn't read TFA but part of the case history was NYT requesting 1.4 billion records as part of discovery and being successfully challenged by OpenAI as unnecessary, and the essence of TFA is advocating for an alternative to the scope of discovery NYT is insisting on, hence the "not rolling over".

Try reading, it's fun!

replies(1): >>45904145 #
freejazz ◴[12 Nov 25 18:45 UTC] No.45904145[source]▶
>Discovery isn't binary yes/no, it involves competing proposals regarding methods and scope for satisfying information requests. Sometimes requests are egregious or excessive, sometimes they are reasonable and subject to excessively zealous pushback.

There is a court order that OpenAI must produce these documents. OpenAI litigated this issue and lost. I'm not sure what point you are trying to make. The court decided the documents were relevant and they must produce a subset of them. Rather than immediately complying, they went and posted this BS "article".

>Maybe you didn't read TFA but part of the case history was NYT requesting 1.4 billion records as part of discovery and being successfully challenged by OpenAI as unnecessary, and the essence of TFA is advocating for an alternative to the scope of discovery NYT is insisting on, hence the "not rolling over".

I don't think you read TFA.

>Try reading, it's fun!

Lol, rudeness aside, you are apparently poorly informed. No doubt it is because you are relying on OpenAI's telling of the events and not actual reporting on the events. Btw, yesterday they were ordered to produce 20m redacted logs. You keep going on about the original discovery request, but that's not what the issue is and it's not the issue OpenAI lost on that they are now crying to the public about.

Also btw, I saw you posting in other comments that OpenAI needs to figure out how to anonymize the data. You probably don't realize this, but OpenAI already represented to the court that the data was anonymized and now are just using this as another delay tactic. Something about "reading being fun". I'd agree. Still, it does depend what you read. Try reading some more!

replies(1): >>45905181 #
1. glenstein ◴[12 Nov 25 19:38 UTC] No.45905181[source]▶
Meanwhile back in reality, as of today that order is being challenged, and challenging the scope of an interlocutory order in discovery is a normal thing and part of a coherent legal position. So I don't know why you're pretending you don't understand what it means not to "roll over like a beaten dog" in response to overzealous discovery.

>I don't think you read TFA.

It was in TFA. If you don't like their number which characterizes OpenAI's interpretation of what an earlier proposal required, the 20 million proposal was selected over the NYT's 120 million records proposal, which demonstrates the same point about fighting to narrow scope. So I still don't understand why you think the concept of challenging the scope discovery is somehow too mysterious to comprehend.

>You keep going on about the original discovery request, but that's not what the issue is and it's not the issue OpenAI lost on that they are now crying to the public about.

Yeah, because I was replying to a comment about that issue and I'm remaining on topic.

>Also btw, I saw you posting in other comments that OpenAI needs to figure out how to anonymize the data.

You're actually right! My mistake. I guess this makes it make sense to pretend you can't understand why a company would ever push back against the scope of discovery.

replies(1): >>45905500 #
2. freejazz ◴[12 Nov 25 19:56 UTC] No.45905500[source]▶
>Meanwhile back in reality, as of today that order is being challenged, and challenging the scope of an interlocutory order in discovery is a normal thing and part of a coherent legal position. So I don't know why you're pretending you don't understand what it means not to "roll over like a beaten dog" in response to overzealous discovery.

That's all OpenAI's argument and not coherent with regard to the facts. That's why they lost. I'd be willing to bet you they lose again. The rest of what you post is just verbatim their side, without any real analysis w/r/t to the facts (again) and I find your responses to this article to be a bit ridiculous in that regard. No less while you criticize others for pointing out OpenAI's incredibly self serving, smarmy, BS about privacy that they otherwise do not actually care about.

Again, this is not an issue of user privacy. OpenAI already represented to the court that they could anonymize the logs and that they did anonymize the logs (something you repeatedly fail to acknowledge while ranting that I didn't read the "article"). The issue is that OpenAI does not want to produce these logs because it will demonstrate that they are wrong. If you're gullible enough to believe otherwise, sure, but it certainly doesn't warrant the ridiculous attitude you bring to communicating with others here.