00:00Competition between the U.S. and China is pushing the development of more powerful artificial intelligence, despite growing concerns over
00:08the risks.
00:09That's the view of AI safety researcher Dr. David Kruger.
00:13He says fear of falling behind China is being used to justify developing systems that he believes are not yet
00:20safe.
00:22We know that what they're doing isn't safe. They tell us that.
00:26And independent experts are also warning about that.
00:30You mentioned safeguards earlier, but the reason I want to emphasize what we need is not safeguards.
00:36We need a moratorium. We need to stop building more powerful AI is because we don't know how to do
00:41this safely.
00:42So the project is inherently dangerous at the moment, and that makes it incredibly reckless to keep going.
00:52And, you know, the argument for continuing this project, continuing to develop more powerful AI for a while now has
01:00been in the United States, well, if we don't do it, somebody else like China will.
01:04And, you know, as long as they're going to do it, we may as well do it as well, even
01:08if we know the whole thing is very risky and could literally lead to the end of humanity.
01:13But that's not like a reasonable response if there is the possibility of international coordination.
01:20And I don't think we've seen serious efforts internationally to shut this project down.
01:26But I'm optimistic that we might start to soon, given how dramatically and quickly the public conversation has shifted recently.
01:34Washington and Beijing have agreed to set up a channel for handling AI-related incidents with plans for further talks
01:42in November.
01:43And on Tuesday, U.S. President Donald Trump announced a voluntary AI safety agreement with major technology companies.
01:51The company said they will commit to conducting internal safety checks and independent audits with board committees overseeing their efforts.
01:59But Kruger said those measures do not go far enough.
02:03He said countries need to call for an international moratorium on the development of more powerful AI systems until the
02:10safety risks can be better understood and managed.
02:14So first off, we absolutely need to not just slow down, but actually pause AI development.
02:20We need to stop building more powerful AI systems because we don't know how long we need to figure out
02:27all of the technical and social problems.
02:29It looks like it could take a long time and we can't count on solving these problems on a deadline.
02:35So we need an immediate, indefinite international moratorium on frontier AI development.
02:41What would that look like?
02:42So I think the hard part of this is the international coordination part.
02:48I think domestically, it's pretty straightforward for the government to just tell the companies you're not allowed to keep developing
02:55more powerful AI systems.
02:57And you can have whistleblower policies.
03:00You can have government or external independent auditors, third parties at those companies monitoring their activities to make sure they're
03:07not doing it.
03:08But that's straightforward and just a matter of political will for international coordination.
03:14You need to have an arrangement where countries like the U.S. and China can be confident that their rivals
03:22are not continuing to develop more powerful AI in secret.
03:26And that is harder, but I think it's still doable.
03:30I think it's still doable.
03:30I think it's still doable.
03:32I think it's still doable.
Comments