On the sidelines of the e-py days (pronounced “hippies”?), I had the pleasure of answering Nicolas Rébières’ questions from La République des Pyrénées, about what we mean by “Trustworthy AI”. The discussions were truly enriching, and the report published in La République des Pyrénées must have been a real challenge for Nicolas R. We discussed the risks of malicious use of AI as well as societal risks. It was one of the few (long) interviews where we even mentioned the fantasy of the tool turning against its creator! It was a fascinating discussion and, I hope, a small step towards calming (some) minds and trying to see the tremendous opportunities behind the new world we are building. It is up to us to make the right choices to use this tool in the direction of progress we want to achieve.
“One of the problems we have is that the progress made in recent years comes at the expense of our ability to prove things.”
It is also important for us to be aware of the limitations of these seemingly omniscient and powerful tools, giants with feet of clay. As long as no responsibility or guarantees are required, they are remarkably effective (and very impressive). But as soon as the slightest error is unacceptable - or requires questioning responsibility - then the limitations are strong.
Find this interview by Nicolas Rébières on the following link (behind a paywal (and in french), sorry)