> The harms engendered by underestimating LLM capabilities are largely that people won't use the LLMs. Speculative fiction about superintelligences aside, an o

> The harms engendered by underestimating LLM capabilities are largely that people won't use the LLMs. Speculative fiction about superintelligences aside, an obvious harm to underestimating the LLM's capabilities is that we could effectively be enslaving moral agents if we fail to correctly classify them as such.