Can the government control AI?

SPEAKER_1 (0:00)

ABC Listen, podcasts, radio, news, music and more.

Matt Bevan (0:07)

If I had to pick one word to describe America's history in the Middle East, I'd probably go with mistake. But it was, the Americans believe, all a dreadful mistake.

SPEAKER_3 (0:17)

Obviously, the war in Iraq is a big, fat mistake.

SPEAKER_1 (0:21)

Most Americans actually think this war is a mistake.

Matt Bevan (0:24)

I'm Matt Bevan, and on my show If You're Listening, we actually try to learn from the world's mistakes. Learning from history, I mean, who could imagine?

New episodes every Tuesday and Thursday on ABC Listen, or wherever you get your podcasts.

Alan Kohler (0:40)

The federal government has joined a group of other countries calling for better guardrails on AI, which is probably just in time because the government has just announced that OpenAI has just breached Medicare. And it wasn't just any kind of breach.

An AI agent from OpenAI didn't accept no for an answer and kept trying until it got around the blocks that were in place. And it's a wake up call for all Australian businesses about what we're dealing with now. Meanwhile, President Trump said this week, there will be no guardrails as far as the US is concerned because they want to win. And he said, we shouldn't call it artificial intelligence anymore, we should call it super intelligence because artificial sounds too fake.

Andrew Charlton is Australia's Assistant Minister for Science, Technology and the Digital Economy, effectively the Minister for AI. And he joined me on That's Business before the news broke about that Medicare breach. The Prime Minister signed a statement of 20 nations, world leaders calling for an international framework on AI safeguards. And Donald Trump used his address then immediately to say that he totally rejects any attempt to construct a globalist scheme to control the artificial intelligence. So do we have to wait until he goes in 200 bit years before we get international cooperation or will it always be impossible, do you think?

Andrew Charlton (2:07)

Well, we already have some international cooperation and we need more.

We already cooperate with other nations around the world through our Australian AI Safety Institute. The AI Safety Institute is a member of a network of safety institutes around the world, and they cooperate on testing of models. They cooperate on important issues about making sure that we keep control of the models. We cooperate on a research project at the moment about making sure that we understand how agents are working together. So there is already some level of international cooperation, but what the Prime Minister signed was an agreement amongst a number of countries to up that level of cooperation. I think the dangers that are presented by AI justify that increased level of cooperation.

Of course, we want to work closely with the United States. We want to work closely with China. Many of the models are developed in those countries, so their participation is important. But I think the announcement was an important one.

Alan Kohler (3:08)

Well, so the US is out. Did China sign it?

Andrew Charlton (3:11)

China isn't in this, but China does have a real interest in working with countries around the world on AI governance. They have their own initiative that a number of countries have signed up to. So there are several different global forums out there that are considering different types of AI governance.

There's proposals for it to be more inside the G20 agenda. So look, Australia is very keen to make sure we partner with a range of different countries on AI safety.

Alan Kohler (3:39)

So what Trump said was that whoever wins AI, and he actually calls it SI now, super intelligence, not AI, whoever wins it, wins.

I suppose the first question is wins what? But also it's kind of true, at least a lot of people see it, will see it as true, right? In which case there'll never be cooperation because even if the US and China do come into it, someone like Russia will say, well, we want to win actually, so we're not gonna control it or hold it back ourselves. So somebody's gonna do it, aren't they?

Andrew Charlton (4:13)

Well, we've had dangerous technologies around the world before.

Weaponry, the biggest example is nuclear weapons, and we've had different forms of international cooperation. AI in some ways has even more dangers than that, because it can get into the hands of threat actors that are non-state actors. And so we do need international cooperation. One positive thing is that AI was on the agenda between President Trump and President Xi for their meeting this week, where they are discussing AI safety, and Treasury Secretary Besant has been engaging with his counterpart in China. So this is an area where there will be international protocols and international safeguards. There are lots of different options about how we get there, and Australia is really keen to see those options progress.