RE: https://mastodon.social/@arstechnica/116364473106973878

When I talk to people who routinely use #AI they often dismiss the (often high) error rates because "AI saves me time when it does work". But if your car broke down every few days, would you still shrug it off because "it saves me time when it does work"? How much of this much hyped increase in efficiency is just users offloading the responsibility to check and fix AI errors onto someone else further down the line, or simply not caring about those errors at all?