7 ms·
Make it so it replaces the comment with some AI slop that's not easily filterable but utterly useless.
by hoseja 3y ago
Make it so it replaces the comment with some AI slop that's not easily filterable but utterly useless.
- gorbachev 3y agoI think the best way to sabotage LLMs trained on Reddit data would be to post something on topic, but straightup wrong, in some other way misleading or with subtle inaccuracies that would cause LLMs to produce bad results in ways that are hard to detect. Use proven information warfare tactics.
- timbit42 3y agoWhy would they import posts or comments with negative feedback?