AI trust demands tools built for professional applications, says Thomson Reuters

Consumer-grade AI tools, such as chatbots and generative systems, are not suited to professional applications and workflows, yet are being widely used within them, leading to poor results and growing mistrust in the technology. That has been a consistent theme in recent reports. Among them, MIT’s finding that 95% of generative AI projects fail to deliver measurable benefits made headlines this autumn and even impacted share prices, though it was based on a small survey base.

Less well reported was research by educational platform Udacity, which found that 75% of the 2,000 US white-collar workers it interviewed regularly abandon AI sessions because of errors and hallucinations, while 45% mistrust colleagues’ work if they know it is based on AI.

That is hardly a vote in favour of tools that, as a matter of routine, are described by their salesmen as having PhD-level intelligence.

Meanwhile, copyright lawsuits over how ...