Post ASk5Eq6roKiPR98goC by hpridham@mastodon.world
 (DIR) More posts by hpridham@mastodon.world
 (DIR) Post #ASk5Eq6roKiPR98goC by hpridham@mastodon.world
       2023-02-16T16:17:00Z
       
       0 likes, 0 repeats
       
       Probably an over simplification, but to address all these #chatGPT prompt injection hacks, why don’t #microsoft #openAI run two instances of the AI in tandem? One sandboxed from the prompts, only reviewing the other’s outputs against the content rules?