Read a post about how people are seeing cascading hallucinations in some organisation's documents because of multiple AI agents running over the same errors again and again.

Lol.

#ai #hallucinations

I read about Karpathy's LLM Wiki and realised that this issue would happen to that too eventually. The operator will get tired and just accept multiple cascading errors because human brains arent designed to think at that scale. Then it slowly gets worse.

@Chloeg

I hope it ends up costing that organisation more money to fix. At the end of the day, the finances won't lie about AI's performance.