00:00Yesterday, we heard probably the most striking warnings on artificial intelligence to date.
00:05AI is killing us all. This is particularly noteworthy because these warnings are coming
00:11not just from outside critics, but from AI safety researchers and insiders who argue that
00:17increasingly powerful systems could eventually become impossible to control. Do I get your
00:24attention? Joining me now to discuss this is Andrea Renda, Director of Research at the Center
00:29for European Policy Studies here in Brussels. Good morning. Good morning. So let me start with
00:35a blunt question. Is this science or science fiction? Do I have to be worried? Do we all have to
00:41be
00:42worried? I think it's science, but we shouldn't be worried necessarily if we keep all this under
00:49control and we keep it public and open. The thing is, it's not the first time that we build something
00:54that can kill us all, right? We have the atomic bomb, we have several powerful technologies.
00:59It all depends on how we use them. And most importantly, we don't have to confuse this
01:04with an announcement that superintelligence is coming, because there are many ways we can all
01:10be killed with stupid technologies if we program them in the wrong way.
01:15I guess the fundamental question seems to be this. Are we building machines that we may
01:21eventually be unable to switch off?
01:25It depends on what kind of goals we give to these machines and what kind of powers, right?
01:29We can have a software which is self-contained, but a software that governs all our critical
01:33infrastructure, the railroads, the telecoms networks and everything, and has the ability to
01:39dodge some of our attempts to stop, especially if programmed to achieve that goal at all costs
01:45might be difficult, might be the genie out of the bottle. But it's a very, very stupid way of
01:50programming AI to give it powers that then cannot be controlled. So we have to keep this in check.
01:57What would have to happen for you to say, stop, AI development must not continue beyond this point?
02:05It's a very difficult question. I think, for now, we don't know enough what is going on inside those labs,
02:11because they are private. And imagine, you know, research on nuclear power happening behind private labs
02:17that no one can control, no one can observe, and people coming out of those labs and telling you,
02:22oh, listen, we're about to kill all humanity, but you don't know how to control that.
02:25I think we just have to open up and audit much more what is going on inside those labs and
02:30invest in public research on this.
02:32I guess we just wait and see. Andrei Renda, thank you so much for coming on the program.
Comments