Skip to playerSkip to main content
  • 15 hours ago
CGTN Europe interviewed Neil Lawrence, DeepMind Professor of Machine Learning at the University of Cambridge.

Category

🗞
News
Transcript
00:00So just how worried should we be? Professor Neil Lawrence is from the University of Cambridge.
00:06PHP, this example, will be one of the cyber security threats we face, but there's going to be many more
00:13coming up.
00:14How worried should we be? Well, we should be concerned as a society, is what I'd say.
00:19Should we be fearing for our lives as we go about our daily business?
00:23Well, no, unless we're particularly paranoid people.
00:25So put this cyber security crisis into perspective, then, if you would.
00:30I mean, just how serious is this chapter?
00:33Well, so to give some context, these cyber security problems are typically caused by a sort of breach in the
00:41so-called data software separation.
00:43So what do I mean by that?
00:45Well, typically we have computers that are programmed by programmers and then people put their information into the computer and
00:53the program processes that information.
00:55Now, the breach I'm talking about is sometimes a flaw in the system means that someone putting in data can
01:02turn that into a program.
01:04And that's very serious because that means they can access other information that might be on the computer, including our
01:10personally private information.
01:12Many of the cyber security breaches, in fact, almost all of them are of this form, this trick.
01:17Now, this is a problem that's been known about for a very, very long time, and many software systems have
01:23this problem.
01:23The risk we have now is that with AI agents, people are able to discover these problems more easily.
01:30But, of course, that means that we're able to fix them more easily.
01:34So it's a sort of it's like an arms race between trying to fix these problems before bad actors find
01:40them and exploit them.
01:41Is this the end of this chapter?
01:44Have we now all learned the lessons and we can move on?
01:48No, this is ongoing.
01:50I mean, I think if we think about millennium bug, this is like a million millennium bugs all over the
01:54place that we constantly have to be vigilant about.
01:57But these problems are what I call type one cyber security problems.
02:01They're problems that have been known about for a long time.
02:03It's just the tools for exploiting them and repairing them are getting better.
02:07So improvement in the weaponry, but the battleground is known.
02:11What we can also expect to see, which is also disturbing, is as people deploy more AI agents, that will
02:17lead to new types of cyber security problems.
02:20Because the AI agents themselves, in effect, breach this data and software boundary I was talking about.
02:27But they breached it on purpose and that's going to lead to a load more headaches to come.
02:32And this is going to be panning out across the next decade.
02:35If I'm a customer of one of these companies compromised, I mean, would I be right to feel rather angry
02:42that my personal details have been perhaps compromised, perhaps relatively easily?
02:49Well, I wouldn't say it's these companies' fault in some sense, unless they fail to patch the software rapidly.
02:57So they need to have security teams that when they're aware of the problem, it's like being made aware of
03:03the hole in your castle wall that someone's intending to attack you through.
03:07Now, if you just sit there in your banqueting hall, ignoring the fact that there's a problem, then that's negligent.
03:13But if you rapidly move to try and patch that hole and secure it, which is what you should be
03:20doing in these circumstances, then those companies aren't really negligent.
03:23And this is more a sort of fact of modern cyber life.
03:26And a bit like, you know, sometimes there are traffic accidents and sometimes those accidents are not really the fault
03:32of anyone.
03:33They're just the way of things.
03:34But we've had traffic laws for a long time and a lot of ways of preventing those traffic accidents.
03:39I think the problem with these cybersecurity problems is they're new and emerging, and we don't have all the best
03:44practices in place to deal with them.

Recommended