Searching his "globally unique name" yields 4800 results.

Good luck with that.

Getting 5580 results now oops. He's got a domain with exact username with his resume which has full name and more, is he serious haha.

If that is truly the case then the original post does seem like trolling. I highly doubt anyone would ever honestly describe several thousand as 'globally unique'.

This is victim blaming. Whether or not he could have been more careful is not an excuse for GPT-3. Illegal behaviour still should be (1) illegal even if the victim could have done more.

(1) I seem to remember a court case somewhere on the planet in the last months where lack of resistance was deemed indicative of consensual intercourse. Which is not even remotely acceptable. But I digress.

What illegal behaviour? GPT-3 is only spitting out public data. Its like blaming someone taking photo of a road for the photo including some random person.

To the best of my knowledge: GDPR has no exception for public data. If you have PII, you need a reason. Consent is a reason; "needed for the requested business" (eg. home address for online shops) is a reason. "Having fun with autogenerating sentences"... doesn't seem like one.

It's one thing for someone to see my username on a gaming forum, search for it, find my github, pick a repo, click on the license, and find my name there. I'm ok with that, I feel like it's a high enough barrier for casual trolls and bots.

It's another different thing for my name to be auto-completed by the most popular, publicly available language model. That I'm less ok with, and I'm sure other people will find absolutely despicable.

We have GDPR and Right to Be Forgotten for a reason.

If someone wants to find out your identity, they're not going to turn to GPT-3, they're going to do the Google search. With that in mind, I don't see how GPT-3 turning up the same stuff that Google does when given your username is a threat to you.

If there's a prompt out there that doesn't contain your username but does spit out your full name, that's a bigger concern.

EDIT: Oh, and as far as bots go: I really can't imagine someone coding their bots to rely on GPT-3 for personal info. GPT-3 doesn't have an "I don't know" answer. In your case it might turn up something useful, but for most people it would turn up nonsense that is indistinguishable from something useful. It's far more reliable and most likely cheaper in the long run to just buy the data.

My name is Matt S Trout. I regularly receive snail spam addressed to 'Ms. Trout'.

Buying the data is quite possibly more cost effective but 'far more reliable' is, I suspect, an overestimate of the quality of the data available to buy.

Well it hugely matters if you have since used the right to be forgotten since the model was made, or if PII was illegally made available and has now been removed.

Unfortunately, your post has led to an huge Streisand Effect—there is now increased interest in knowing who you are.

Just take a look at Google Trends for your username: https://trends.google.com/trends/explore?date=now%201-d&q=Bo...

If I were you I'd do the following:

1. Email dang ([email protected]) to ask how he can help with damage control.

2. Stop commenting under your now de-anonymized alias. Any future posts on this topic should be from a freshly minted HN account.

3. There are still quite a few open source repos under your GitHub handle that contains your full name in the LICENSE.txt, so you might want to strip those out as well.

Good luck with getting OpenAI to extricate your full name from their model.

The fundamental mistake here was posting under the same name he'd been using all over the internet for at least 13 years!

The larger question I think is what happens with OpenAI when it comes to information specifically prevented from publication by a court order - for example the new name of Robert Thompson. How can an AI be held in contempt of court?

What about when AI puts an end to witness protection as it works out who the new identity is. Between behavioral and photo matching it should be quite easy. What happens when it gets it wrong?

Thank you very much for your concern, but my anonymity was very very casual, and it doesn't bother me much that it's now lost.

I was hoping to create a conversation about privacy in language models and it's slippery slope, and in that it was partially successful.

I think what the conclusion, though, is that the loss of privacy to language models pales in comparison to the loss of privacy to the rest of the internet.

The language model is just a different way of organizing the information that is already out there. It doesn’t create or expose any new information, and it’s ability to correlate information with other information isn’t superior to the other, existing, ways. If someone is attempting to find out the real person behind your username, a google search still works better than typing your username into the language model.

The real lesson (and one that predates language models like GPT), is that we can no longer rely on the strategy of being ‘lost in the crowd’ to preserve our privacy. When you took all of those actions that connected your username to your real name, you probably thought, “this is just one little obscure thing, no one will ever find it.” Of course, Google finds everything.

This is completely unrelated to things like GPT.

> The language model is just a different way of organizing the information that is already out there. It doesn’t create or expose any new information, and it’s ability to correlate information with other information isn’t superior to the other, existing, ways.

At some previous logistics company this was something I questioned regarding a new tool for managers that could more easily track driver performance as opposed to digging in many old systems, taking notes and making inferences.

Because the new tool made it easier, even though the data is the same, the company's lawyers were very skeptical of this in relation to the driver's privacy.

There is something grey about this I think. Ease of access do matter to some extent, even though my engineering brain says the data is the same so it doesn't matter.

I actually completely agree with this.

A lot of our historical privacy has been based on practicality. We don't worry about someone monitoring us all the time in public because it would be expensive and exhausting to follow us around all the time, and there is no way to follow everyone.

Computers change this. Computers are great at doing repetitive, boring, tasks forever. A person couldn't watch every camera 24 hours a day, but a computer can.

It is absolutely something we should think about.

The problem with using your example to start this conversation is that there's good reason to believe that GPT-3 memorized your name—username relationship precisely because you were so causal in linking them. There was so much data out there that it ended up embedded in the model.

It's hard to get worked up about privacy in language models when the big scary case in front of us is one where the person in question admits they never cared much about maintaining anonymity.

Yeah and the conversation you started is "people who care about anonymity aren't serious about it."

That's kinda the point though - the argument being presented is AFAICT "people who aren't that serious about it should still not be randomly outed."

I'm not convinced that's practically tenable these days, but on a moral basis I endorse it and would present Scott Alexander as an example of why.

Okay, but Scott actually had a reasonable argument for why the level of anonymity he wanted was important. Scott was not trying to start a conversation, Scott was trying to be anonymous. That's why he's such a good example.

But it looks like it's in all sorts of places - how do you think the language model ended up thinking that was a good autocomplete?

There appears to be a name on a YouTube channel too that doesn't even need any additional steps (ie the jumping into repos and then licence files you mention above)

Putting that aside, why is this such a concern? it's just a label. It would be another thing if you had something more meaningful being revealed (eg current physical address) but username / real name alone is generally not that big a deal and it's plausibly deniable (ie there could be plenty of other username / real name pairings that are valid)

Anyway, this comes across rather like the Streisand Effect

your username contain part of your actual name?…

…Not to reveal information here (counter to your goals) but there was no way this thread wouldn’t motivate people to see how hard it is to find your name. But I was surprised to see the username contained a word from your real name considering you’re concerned about revealing personal information.

Making a new online identity is pretty standard practice these days if you’re worried about this sort of this. Especially if you read Grugq’s stuff about modern opsec (just google his name).

Sure, my username is part of my actual name. I can still wish for a modicum of privacy.

If you find that so strange, read up on Scott Alexander[1], who deleted his entire blog when the New York Times threatened to publish the third and last part of his name.

I'm trying to keep basic hygiene, like not having *Hacker* News on the first page of results when somebody googles my real name.

[1] https://en.wikipedia.org/wiki/Slate_Star_Codex#New_York_Time...

IIRC for Scott it was a professional problem: a psychiatrist's patients are not supposed to know anything about him outside therapy.

How do you know it's your name? Maybe it's someone else with the same name as you?

Because of my home country naming standards, I have two unusual last names. There's no one else with the same name as me.

you established an online identity that is apparently persistent across many different websites and also intentionally linked it to your personal identity

this has nothing to do with AI or GDPR. the places you gave this information were public. when you do this there’s no way you “ try to hide my real name whenever possible, out of an abundance of caution.” you’re doing all the things any basic infosec lesson would tell you to not do

What prompt did you use on GPT-3? I tried a few and it never revealed your actual name.