Hacker News new | past | comments | ask | show | jobs | submit login
Meta’s new AI chatbot exposes anti-semite, homophobic views (cybernews.com)
2 points by giuliomagnifico on Aug 8, 2022 | hide | past | favorite | 2 comments



It's the old sentiment of Garbage In, Garbage Out. You train a bot to be like humans, it'll be like humans, and humans (especially when anonymous, like on the internet that these bots are trained from) are awful.

(And let's be clear: GIGO applies to humans as well. How many people are pieces of shit in large part because they were raised that way or otherwise learned to be from bad data?)


> While the chatbot has triggers when it considers the topic potentially unsafe

This sounds like a pretty boring chatbot.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: