About this programme
Can you trust AI once it has been right ten times? An AI founder says that is exactly when the 11th mistake slips through. It has a name: automation bias.
Short answer: treat AI like an assistant and check its work every time, however good it has been.
Fakhar Anwar Safeer asks Rezo Tateishvili, who works with AI every day, what genuinely worries him. Rezo says the danger is trusting AI because it has done good work many times, then missing the one mistake, and having to fix it and go back and forth again. That is why he doesn't want to become too dependent on it: he treats AI like an assistant, telling it exactly what to do and always checking the work it has done.
The facts: • Automation bias is the tendency to over-rely on an automated system and stop checking its output (Parasuraman and Manzey, Human Factors, 2010).
Sources and further reading
Sources mentioned in the programme will be listed after they have been checked and attributed.
Transcript
A transcript will be published only when an accurate version is available.
