4 ms·
I see several converging factors 1) It’s impossible to distinguish between human and AI. Breaking a captcha is a resume building weekend project for any dedica
by swagasaurus-rex 3y ago
I see several converging factors
1) It’s impossible to distinguish between human and AI. Breaking a captcha is a resume building weekend project for any dedicated hacker.
2) The turing test has thoroughly been solved. There’s already a trend of accounts on reddit that reply with helpful but obviously ai generated content to farm reputation.
3) Spam and troll content is still an unsolved problem that requires a dedicated team to solve.
4) Communities themselves can be hostile, especially when the majority adopt black and white thinking.
5) Vetting new accounts is mostly done by verifying email or phone, and doesn’t filter out bad actors.
Some of these problems are solvable. The AI ones seem like a tall hurdle. There are some verification methods like video that AI struggles to replicate but it will soon be intractable to distinguish between human and AI online.
- red-iron-pine 3y ago> 4) Communities themselves can be hostile, especially when the majority adopt black and white thinking. many of these communities will inevitably fall into that thinking, even when they start out as a more open forum, due to selection bias, us-vs-them / in-group-out-group, and a tendency for more controversial, angry, or loud posts to get more views.
- fragmede 3y agoAnd even if I’ve met someone briefly, who’s to say they haven’t wired a bot into slack to speak for them? At what point does it cross the line? I fully admit to using ChatGPT to help organize my thinking, but that seems benign compared to wiring up a bot and have it fully shitposting as me.
- fragmede 3y ago5) vetting new accounts for some of the more private communities is invite only, so you’ve got to know somebody. Limits the growth but makes sure there are no bots.