Paranoid types who tend to overthink these things might want to sit this one out…
In this blog post, the author reports that apparently the new Claude can identify authors of unpublished text with very high accuracy: I can never talk to an AI anonymously again
I wrote up my thoughts here but they amount to “well, that sucks.” Curious to hear others’ takes.
Is your site anonymous? Why/why not?
Are you worried about this?
Was this inevitable?
To what extent does internet anonymity resemble age verification as a policy/ethics issue?
Pseudo anonymous for the moment. I grew up with an internet where we were encouraged to avoid revealing personal information - and had SO much fun with role play in chatrooms! - sleepwalked into years of being less careful, and have come out the other side incredibly cynical about how safe my information is online.
I’m moving towards more private platforms, but I know of at least one instance where my first name is still on my website, and I’m sure some family member’s names are likely to be too. It’s likely not hard to work out who I am unfortunately.
Not actively, if only because I’ve internalised the mantra that nothing online is ever really deleted. I have nothing to hide - I have vulnerabilities, embarrassments and lapses of judgement like the best of us - but at the same time that doesn’t mean I sleep well with the thought that anyone can have access to that information.
Achingly, depressingly so. I’d love to be part of the anti-AI movement. I don’t agree with the use of AI in so many respects and I don’t use it on my website. However, the reality is, I do use AI. I don’t use it to the extent many people seem to and I really don’t see it as a magical bullet, but it would be disingenuous to claim that I don’t find it useful or to claim I don’t use it when I’ve been tempted back twice already. I am absolutely alarmed at the rate of change when it comes to AI and that it’s now in every conversation. I actively try to avoid it at work, but I’m still a work in progress.
Imho they massively overlap in ways that I do find hard to reconcile. To me they are both about identity and safety. I believe in the right to express yourself openly and live as your authentic self, but I don’t believe that the state should have a say in what that looks like. I believe in the right to safety and security of the individual and the wider community, but I don’t believe our current governments and police institutions are the answer to that. I don’t believe big tech has my best interests at heart, despite its convenience, and it is all the more worrying when it overlaps with government policy.
I believe people should have the right to remain anonymous. At the same time, I don’t believe anonymity is a free pass for hate speech and harassment. I just don’t think our current institutions are the right framework to tackle that. I also have a massive slice of privilege that allows me to say so openly.
You could argue that proving who you are is not the same as proving your age, but I don’t believe that. We leave fingerprints on everything and even the best intentions are often misguided, harmful or twisted to suit others’ interests.
I agree with everything you wrote. I am also shifting a lot of my attention to private forums/group chats but even this feels like it gives more control to the platforms. For example many of my best “posts” are whatsapp messages with a specific couple of friends and there is no easy way to dump those to a flat CSV, back them up, etc.
As someone who fully doxes themselves online, I believe that anonymity deserves to be an absolute right on the Internet. It does not make people behaviour more poorly, the most rancid comments I’ve seen have been from public accounts with full names and faces on Instagram and Facebook.
I’m very worried, and the truth is a lot of anonymity has already been compromised. There’s a multitude of ways to fingerprint someone online. It was inevitable the way every other stripping of rights has been, through oppression, fatigue, complicity and apathy.
Also, I tried out what Piper tested and it didn’t work with me, but that’s probably because I’m nowhere near famous enough lol. I think, luckily, the amount of resources that would be required to have a writing sentiment profile of everyone via LLMs would be too costly with the bubble already beginning to pop.
This reads like a stealth ad for Claude. Average subshit article (substack bad) lol. The identifiability of a journalist with a decade long body of professional work is going to be more prominent in AI training data than text by random people.
If I’m to panic, I’d rerun the test with randos. But I’m not the one paying for LLM SAAS lol, so I don’t have access to the models used here.
De-anonimising text is something humans do, manually. And its very unreliable! I don’t trust an LLM could be more reliable.
LLMs can’t even reliably detect LLM generated text. Having a way to know who wrote every post is simply not something I think will happen.