AI boom
AI boom© Provided by The Telegraph
Dan Snow AI apocalypse gif
Dan Snow AI apocalypse gif© Provided by The Telegraph

At 5:29am on July 16 1945, the world entered the nuclear age. Deep in the New Mexico desert, scientists watched the first test of “the gadget”: a nuclear bomb with a 13lb radioactive core. This was “Trinity”, the climax of the Manhattan Project, one of the most expensive research and development projects ever undertaken. The men who had birthed this atomic weapon watched through protective goggles, shielded in bunkers 10km away from where the gadget had been hoisted up a 100ft tower to mimic the effect of being dropped from a bomber.

After a rain delay, they flicked the switch. The desert sand beneath was instantly turned to glass for hundreds of metres. The observers were overwhelmed by the light and heat. The mountains were illuminated brighter than by any desert sun, the heat was like “being in an oven”. The intensity of the blast stunned even the boffins. Robert Oppenheimer, the theoretical physicist and lead scientist, remembered later that “a few people laughed, a few people cried, most people were silent”. He was reminded of Hindu scripture, “Now I am become Death, the destroyer of worlds.”

For at least one of the scientists there was a modicum of relief. Enrico Fermi had speculated that there was a chance the atmosphere itself would ignite and incinerate the entire planet. Just weeks later, two atomic bombs were dropped on Japan, tested on human flesh rather than desert sand.

I had a less spectacular epiphany this week. At 2pm on Tuesday, I saw pictures of Donald Trump getting arrested, resisting police officers, being manhandled to the ground. They were deeply gratifying but also deep fakes. The product of some hopeful creator who could not wait for the day on which justice does indeed catch up with the orange Caligula.

The nuclear revolution was one of the most transformative technologies ever produced, it gave us the ability to destroy life on this planet. The explosion of AI this year may have a similar impact. It will change everything. We just don’t know how.

Through luck more than judgment the world has avoided further combat use of nuclear warheads. Nuclear arsenals have ballooned. The strength of the weapons vastly increased. In the 1960s, the Soviet Tsar Bomba was tested, producing a yield of 50 megatons, the equivalent of 50 million tons of conventional explosives. That’s 10 over 1,500 times the force of the Hiroshima and Nagasaki bombs combined. We have been seconds away from accidental detonations at early warning facilities in the US and Russia, which could have easily led to an accidental launch. During the Cuban Missile Crisis, there was a shouting match on one Soviet submarine between officers debating whether to fire a nuclear tipped weapon at harassing US naval ships. Will the same be found true of 21st-century innovation?

My nervousness about the new technology was not soothed by the letter this week signed by an illustrious group of tech titans, including Elon Musk, who begged the AI companies to suspend all research for a period of six months. They tell us that there is an “out-of-control race to develop and deploy ever more powerful digital minds that no-one – not even their creators – can understand, predict or reliably control”.

I thought of the striking parallel between these people and the physicists who expressed regret and concern after the birth of the bomb. Albert Einstein, who helped to convince President Roosevelt to set up the Manhattan Project, later said “had I known that the Germans would not succeed in developing an atomic bomb, I would have done nothing”. Leo Szilard first conceived the idea of a nuclear chain reaction while standing at traffic lights near the British Museum on a rainy day in 1933 and became a committed opponent of nuclear weapons, organising a group of scientists to call for disarmament. “I feel the responsibility,” wrote Szilard, “of having invented something that could be of immense benefit to mankind, and equally of immense detriment to mankind.”

Nonetheless, this week, the Government sallied forth as the tech savants issued their dire warnings. “We need to have rules to make sure it is developed safely,” said science, innovation and technology secretary Michelle Donelan, presumably while casting around for a few planks to block up the stable door, inside which there is no horse.

Worse still, it’s the wrong stable. ChatGPT is American, as is Midjourney, the engine that built those striking pictures of Trump. The other big player? The Chinese Communist Party, who we cannot assume will willingly submit themselves to Ms Donelan’s rules.

Nearly every pressing issue we face – migration, climate, banking instability, Ukraine and now AI – can only be solved by working internationally. National policy might win an election, but any serious attempt to solve the issues means transnational action. Now.