Interview with Christian Szegedy, Discoverer of AI Adversarial Examples [video]

&gt; Today, he&#x27;s working on formal reasoning and dreams of creating an <i>automated software engineer</i>.<p>How does that work and should we be worried? The article doesn’t touch on this topic beyond the introduction.<p>(Note: I am aware of using genetic algorithms to evolve code to produce a solution to some specific problem. &quot;Automated software engineer&quot; seems to contemplate something more sophisticated than that.)

This is a good interview. I was a little disappointed that it did not dig more deeply into adversarial examples. The compensation was that Szegedy argued that human-like reasoning may be based on relatively simply mechanisms that we just need to find. He cited AlphaGo as a solution that turned out to be much simpler than expected.