[jkuli]

Humans can also make mistakes. This is the first test I apply every time. Could it be that AI is actually more capable than a human? If industry decides than humans are more reliable, they will choose to use humans. Reliability is part of cost-effectiveness, and it's built in to the business decision process.

[theodric]

A human would only make up an answer like the ones in the article if it were a compulsive liar. A human would ideally say "I don't know" or at worst employ the "I'll confirm that and circle back" corpspeak evasion.

[mdp2021]

Another case of badly computed reputational loss.

That context is not of "this more than that" comparisons, but of threshold: the service must be _adequate_.

If you don't have random humans capable of providing the needed service, find them. Do not employ random agents.

[alpaca128]

The average Google user expects LLMs to be perfect and treats their responses as answer from an oracle, not just better than the average human.

It doesn't matter whether it's better than humans, the one thing that matters are the consequences of its widespread use.