2. "Confident but wrong"
We seem to have a little AI theme going on here lately!
In what is possibly the worst combination of results, a newly published study into generative AI-driven chatbots found that they were confidently providing incorrect information on a range of health-related topics that are frequently searched for.
As reported in BMJ Open, the researchers looked at five popular AI chatbots, asking questions about cancer, vaccines, stem cells, nutrition and sporting performance. Very nearly half of the responses were incorrect in some way.
We have more information here
|