Being too nice online is a dead giveaway for AI bots, study suggests
New "computational Turing test" reportedly catches AI pretending to be human with 80% accuracy.
https://arstechnica.com/information-technology/2025/11/being-too-nice-online-is-a-dead-giveaway-for-ai-bots-study-suggests/?utm_brand=arstechnica&utm_social-type=owned&utm_source=mastodon&utm_medium=social
Discussion
Loading...
Post
@arstechnica Considering, I'm working to be nice on the internet. I guess the test will soon tell me I'm an AI bot.
@arstechnica What an insightful article. You have several powerful ideas there.
@arstechnica Well fuck !
🤦♀️
@arstechnica It's that fake-nice thing that all chatbots do. It is a dead giveaway, and doesn't hold a candle to genuine human kindness and compassion.
This reminds me of MGonz - an Eliza bot that fooled people back in the early 90's when only nice people were on the internet - by pretending to be an absolute arsehole rather than giving the usual Eliza responses.
Anyway. Turns out the guy who wrote it is a bit off the deep end these days - but so it goes.
@arstechnica obviously that’s utter rubbish. Being too kind means LLM? What a joke.
@arstechnica So, I need to be an asshole. Otherwise it's a dead giveaway that I'm a bot. Ok, sure...
@arstechnica
Fuck you, ArsBot.
Did you have to do that? I forsee #LLMs (in addition to all the other problems they bring) now starting to emulate nazis and trolls to blend in and work around that detection.
I'm seriosly considering becoming Luddite at this point.
bonfire.cafe
A space for Bonfire maintainers and contributors to communicate
Automatic federation enabled