> Some of the AI-generated output is factually wrong While this is undoubtedly true and a problem that needs to be addressed, it's worth considering that human

> Some of the AI-generated output is factually wrong While this is undoubtedly true and a problem that needs to be addressed, it's worth considering that humans gets things factually wrong too sometimes (intentionally or not). So perhaps a more interesting question is how much more or less correct an AI is than a human on a given task. People talk about this with self-driving cars all the time. Arguably a self-driving car does not need to drive perfectly (not that this is not a good goal), but if it can drive more safely than the average human driver, there's still a significant chance of improving overall road safety.