Comment by mhitza
12 hours ago
Thanks for the providing the details, where I've been just lazy about reading the paper now :))
I'm not a fan of your proposed changes, as they further lock down platforms.
I'd like to see better tools for users to engage with. Maybe if someone is in their Firefox anonymous (or private tab) profile they should be warned when writing about locations, jobs, politics, etc. Even there a small local LLM model would be useful, not foolproof, but an extra layet of checks. Paired with protection about stylometry :D
Mitigations are pretty difficult, I understand it is kind of cool that some websites have really open APIs where you can just read everything. There are some cool apps that used HN data in the past. But I think there should at least be consideration that LLMs are then going to read everything and potentially discover things. Users might have thought this is protected by obscurity, who would read their 5 year old comments?
How helpful would injecting noise and red herring into pseudonymous posts help?
It seems like it would make sense to get in the habit of distort your posts a bit, and do things like make random gender swaps (e.g. s/my husband/my wife), dropping hints that indicate the wrong city (s/I met my friend at Blue Bottle coffee/I met my friend at Coffee Bean), maybe even using an LLM fire off posts indicating false interests (e.g. some total crypto bro thing).
This is probably a good use case for something like OpenClaw. Have it take over your accounts and inject a bunch of non-offensive noise using a variety of personas to pollute their analysis. Meanwhile, you take your real thoughts and opinions underground.