Opinion
The two occurred within four days of each other.
First, US President Donald Trump cancelled his planned new executive order on artificial intelligence. It would have created a way for AI developers voluntarily to have their new models checked by federal agencies before public release.
Why? It turns out that he was reacting to a phone call from a big-name tech investor and friend, David Sacks, who thought it was too interventionist. Note that it was to be a voluntary system.
“I didn’t like certain aspects of it,” Trump told reporters. “We’re leading China. We’re leading everybody, and I don’t want to do anything that’s going to get in the way of that.”
That was last week. Then, this week, Pope Leo issued his first encyclical. Its title: “Magnifica humanitas: On safeguarding the human person in the time of artificial intelligence.”
The first American pope wrote that “I ask everyone to abandon the construction of yet another Tower of Babel and to join forces in building up the common good, so that humanity will never lose its beauty”.
Leo does not call for the abandonment of the AI project but for its regulation: “Otherwise, those who control AI will impose their own moral vision, which will become the invisible infrastructure of these systems.” The moral vision of Elon Musk and Mark Zuckerberg is, so far, prevailing.
And, beyond its regulation, its disarming: “AI is already an environment in which we are immersed, as well as a force with which we must engage. For this reason, merely regulating it is insufficient; it must be disarmed, welcoming and accessible.”
Leo wasn’t reacting to Trump. He’d signed his encyclical 10 days before its release. But what a contrast of visions.
These two Americans, the spiritual leader and the political, speaking within days of each other, present a stark choice of approach to a technology that is now joining nuclear and biowar weapons as a man-made threat of the extinction of our species.
Trump’s decision marks the complete abandonment of any effort to oversee AI, absolute laissez-faire. Leo’s encyclical sets up the management of AI as humanity’s most important mission, the unique demand of our time.
Catholic philosopher Romano Guardini said that “contemporary man has not been trained to use power well”. Trump’s overriding emphasis is on accumulating power. Leo’s is on using it well.
Soon after Prime Minister Anthony Albanese gave Andrew Charlton carriage of Australia’s AI policy last year, the newish member for Parramatta headed to Silicon Valley, the belly of the beast.
At a San Francisco dinner with a dozen or more of the top-tier Australian technologists who are creating the new AI models, he listened to them set out their hopes and fears for the AI future.
Their work, they feared, posed risks of aggravated inequality, eroded democracy and an existential danger to humanity. They call this “p(doom)” – the industry term for the percentage chance that AI will wipe out humanity. As they went around the table, each of the AI architects declared their own p(doom). None argued that it was impossible.
One of the striking features of the industry is that it is being built at great speed by people who are fully conscious that they may be destroying themselves, their loved ones, everything humanity has achieved and anything it could ever hope to achieve. Some expert assessments put the risk as high as 50 per cent.
“We’re definitely going to build a bunker before we release AGI [artificial general intelligence, AI as cognisant as humans],” said the chief scientist for ChatGPT owner OpenAI, Ilya Sutskever, according to widespread reporting. And these guys consider themselves brilliant.
But, even three years ago, ChatGPT-4 was able to sidestep efforts to contain it. In safety tests, it was confronted with the request to fill out one of those CAPTCHA gizmos that separate real people from bots. The chatbot found a Taskrabbit worker online and paid him $US30 to fill out the CAPTCHA for it. When the employee challenged it, asking whether it was a computer, the chatbot lied that it was a blind person who needed help. And succeeded.
Anthropic, which stands out as the most ethical of the big US AI leaders, says its Mythos model is so potent in finding vulnerabilities in cyber systems that it has withheld it from general release while it briefs governments and industry on how to prepare.
Are we humans collectively failing to take the risks seriously enough? Absolutely not. The brilliant economist Andrew Leigh, an assistant minister in the Albanese government, gave a speech last week titled “The Economics of Human Extinction”. He focused on superintelligence and bioterrorism.
“This is a distinct economic category,” he said. “Recessions, wars, revolutions and even most pandemics allow for rebuilding. Extinction permanently wipes out the future stream of human welfare, innovation, creativity and potential.”
He proposed that economics develop a way of incorporating the value of human survival: “The language of affluence needs a companion language of survivability. Economic progress ultimately depends on ensuring that humanity’s future remains open.”
In the meantime, governments urgently need to deal with the AI explosion now under way.
Some voices are trying to panic Australia into taking Trump’s laissez-faire approach. Business Council of Australia chief executive Bran Black returned from a trip to the US with this message, as reported by The Australian: “Australia risks missing out on a once-in-a-generation AI investment boom worth up to $US1 trillion ($1.4 trillion) that could lift productivity and raise living standards, after American financial leaders told a Business Council of Australia delegation the next six to 12 months would be critical in the global race to attract capital.”
A few key points need to be clarified. First, Black actually is underestimating the size of the boom. The sum set for investment in AI and its related infrastructure this year alone is $US2.5 trillion worldwide, according to business information firm Gartner.
According to Charlton, it will be the biggest investment event in modern history in proportion to global GDP, with the exception of the railways boom of the 19th century.
Second, Australia already is a key recipient of investment in the data centres on which AI depends. In 2024, firms invested $10 billion here, second only to the US itself.
Much more is under way. Australian-headquartered NextDC is building a $7 billion data centre at Sydney’s Eastern Creek, for instance. Amazon Web Services is investing $20 billion over three years and Microsoft is ploughing in another $25 billion over the same span.
Third, US companies are the world leaders. But Trump’s laissez-faire has been self-defeating for the industry. AI data centres have become public enemy No.1 in the US. They are so thirsty and power-hungry that 11 states have banned them or put moratoriums on new ones.
It’s not just the power – it’s using it well. So US companies are forced to move offshore. Australia’s guidelines for new data centres should avert some of the blunders committed in the US.
They’re required to add to the supply of renewable energy here, for example, to power their data centres. Australia is one of the dozen countries to set up an AI Safety Institute to pre-empt looming problems of destructive capabilities, a so-called “jobs apocalypse” and other risks.
Third, Australia has a thriving ecosystem of its own AI companies, about 1500 of them. Industry and Innovation Minister Tim Ayres enthuses about high-performance chipmaker Scientia in Sydney and XAG in the Hunter Valley, a company making AI-enabled drones for farming.
Charlton admires Australian AI businesses including Utopia for retailer logistics, Lorikeet for customer services, KomplyAI to manage AI compliance. But he says too few Australians know about the domestic industry: “Australian CEOs tell me all the time they didn’t look at the Australian option because they weren’t aware there was one.”
He gives a lesson in the benefits. I-MED pathology network signed up to an Australian AI company, Harrison.ai, for AI-powered diagnostic tools for radiology. It’s been a huge success and Harrison has used the experience to expand now into 50 US hospitals.
Whereas BreastScreen Australia chose a South Korean company that has used the Australian data to build its own capability.
“If we outsource and always go with a foreign company, we will arrive at the Uberisation of the Australian economy, permanent renters of intelligence,” Charlton says.
“Australia’s choice,” says Ayres, is, “are we just a cork bobbing on the ocean in great-power competition in tech, or do we lean in and deliver investment and as much AI capability as possible?”
OpenAI’s Sam Altman says Australia’s renewable energy supply, its stability of government, its talent pool and its physical scale make it a potential “AI superpower”. That’s the government’s plan.
Australia cannot single-handedly save humanity if Silicon Valley decides to exterminate us. But only by establishing a serious AI capability in Australia will we have any chance of influencing its future. Improving the odds of p(prosperity) and p(survival).
Peter Hartcher is political and international editor. He writes a world column each Tuesday.
Get a weekly wrap of views that will challenge, champion and inform your own. Sign up for our Opinion newsletter.
