3 ms·
A year or so ago I experimented with splitting a user prompt down to a set of "different AI personas" that would each try to approach the user's problem in a di
by vunderba 2y ago
A year or so ago I experimented with splitting a user prompt down to a set of "different AI personas" that would each try to approach the user's problem in a different way and then bubble back up with a master arbiter for consensus.
I modeled it after the concept of advisors from Civilization II. It worked reasonably well though I think it was at least somewhat limited by being constrained to a single LLM (Mistral). It also lit my computer on fire.
- bee_rider 2y agoWhat sort of personalities did you try? A group where some members have grudges against each other and will irrationally poke holes in each other’s plans could be a fun experiment.
- throwup238 2y agoWith multiple groups with external and internal rivalries. The Always Sunny gang versus The IT Crowd.
- vintermann 2y agoI have played Disco Elysium, and can confirm that a bunch of inner voices arguing with each other can be fun.