Technology

Late last year, I published the first theoretical physics paper in which the main idea originated from artificial intelligence – from an AI. And my experience working with the most powerful AI models left me both impressed and wary. The most accurate analogy I can offer is that it’s like something familiar to anyone who has done research: a brilliant colleague who is also unreliable. This colleague can produce deep insights at surprising speed – and then, the next second, make an error that ranges from the trivial to the profound. That tension – capability versus reliability – has shaped how I now use these systems in mathematics and theoretical physics. It is also what will shape how they affect scientific research over the next few years.

By using a sonic weapon in the mission to capture Nicolás Maduro, as Donald Trump appears to have confirmed, Delta Force commandos not only triggered a paradigm shift in warfare, but served poetic justice. When asked whether such a weapon had been used, the President replied: “It’s probably good not to talk about it.” But then added: “Nobody else has it, we have some amazing weapons that nobody knows about.” The following morning, at Davos, Trump said: “They weren’t able to fire a single shot at us. They said, ‘What happened?’ Everything was discombobulated.

For years, retailers have been behaving like needy friends. My phone would ping. “Hi there!” an email would read, “We’ve missed you.” Who could this be? I would wonder happily, before realizing that the warm and loving message was from someone in the marketing department of the emporium from which I’d once bought a couple of pairs of expensive shoes. Emma usually, or Olivia. With the advent of artificial intelligence, though, personalized, over-the-top PR is getting much, much worse. At least in the past you’d needed an actual Emma to send these emails, some girl who’d gone into marketing, typing enthusiastic nonsense all day. There was at least a human who had “reached out” – they always “reach out.” AI is just so needy.

Donald Trump’s desire for Greenland is not just about access to oil, minerals and control of the new strategic and commercial corridors opening in the region. It’s also about data. Specifically, the most important data in the world. For decades, Pituffik Space Base – formerly Thule – in Greenland has been central to US space defense and Arctic strategy. It’s the US military’s only base above the Arctic Circle and their most northerly deep-water port and airstrip. It’s home to the 12th Space Warning Squadron. Its massive AN/FPS-132 radar has 240 degrees of coverage surveying the Arctic Ocean and Russia’s northern coast, especially the Kola peninsula where it has concentrated its strategic nuclear weapons.

In the early hours of Saturday, January 3, Caracas went dark. Power failed across much of the city as strikes and cyber-attacks disabled critical systems. What followed was not a conventional invasion, but one of the most audacious special forces operations in modern history. Within hours, Nicolás Maduro and his wife, Cilia Flores, had been seized from the heart of Venezuela’s largest military complex. No tanks rolled through the streets. No territory was occupied. The operation succeeded not through brute force alone, but because of something far more decisive: overwhelming American dominance of intelligence, networks, surveillance and infrastructure.

The X-native AI Grok exploded in popularity this weekend – as users discovered that its media tab was filled with requests to generate disrobed and scantily clad versions of images of women and children that people had posted publicly. "Put her in a bikini," users asked the AI. Grok complied with these requests freely, with no meaningful oversight or guardrails in place, automatically generating images corresponding to every prurient prompt. The ensuing discourse quickly polarized. On one side were tech nihilists, arguing that this use of AI was inevitable and therefore unsurprising. After all, anyone can already download publicly posted images and manipulate them privately.

On March 6, 1959, nine-year-old Candice "Candy" Rogers of Spokane, Washington, went out after school to sell campfire mints door-to-door. Sweet-natured with strawberry blonde curls and a button nose, she was small for her age. Her mother, Elaine, had one clear rule that Candy must be home before dark. But Candy never came home. Her disappearance triggered a 16-day manhunt involving thousands of people, the Marines and the US Air Force. In fact, three airmen lost their lives on the second day when their helicopter hit high tension cables and plummeted into the Spokane River.  After two weeks of searching for Candy, all detectives could find were her scattered mint boxes.

The foundation of science is based on the humility to learn, not the arrogance of expertise. When comet experts argued that the interstellar object 3I/ATLAS must be a familiar water-rich comet as soon as it was discovered in July, they behaved like artificial intelligence systems: only able to reflect the data sets they were trained on. For decades, the data set that established comet expertise largely comprised icy rocks in the solar system. My counterpoint is simple: humanity launched technological objects into space, so we must conclude that alien life forms could do the same. This possibility must be added to the training data set of comet experts when studying interstellar objects.