FOUR

Thankfully, there are rosier predictions.

A recent paper from the National Bureau of Economic Research, a US think tank, predicted that recursive self-improvement would lead to a technological innovation loop that creates “explosive” economic growth.

But in all honesty these predictions are little more than an educated guess. A few years ago, nobody imagined that AI would lead to a world in which computer coders were more at risk of unemployment than truck drivers.

That makes preparing for it almost impossible.

It is also worth noting that while AI companies are proceeding largely unchecked, is it really that feasible to believe that a group of government bureaucrats sitting around a table could design and execute the implementation of a set of standards that would improve this?

And some experts, such as those at Apollo Research, says keeping superintelligent AI under control may turn out to be “one of the hardest challenges ever”.

The world’s top AI labs are all American, so realistically, reigning in the technology will fall to the White House.

Up to this point, Donald Trump has encouraged AI to sweep across the economy. He has challenged individual states and foreign governments that have threatened to regulate the technology.

But somewhere in the White House, a switch has flipped.

After Anthropic unveiled Mythos, its chief executive was summoned to the White House. So were leaders of Wall Street’s biggest banks, who were told to upgrade their security to protect against AI hackers. Tech bosses were told by JD Vance, the US vice-president, that “we all to need to work together on this.”

Kevin Hassett, Trump’s economic adviser, has said that future AI systems may need to be safety tested and approved before release, in a similar way to medicines.

The change of direction is unmistakable. AI policy has now firmly entered its science fiction era. Strange things are happening and stranger things yet will happen soon.