AI is like the first car: it will revolutionise a world that has not yet been built for it

We badly need to talk about AI, but at the moment, it’s difficult to have a grownup conversation about it. That’s because we find ourselves trapped in what the Johns Hopkins academic Henry Farrell calls the “AI Fight Club”: bout after bout of “ritualised combat ... between two highly stylised positions”.

Position one is that AI will change everything, and lead to artificial general intelligence and a post-singularity paradise. Position two is that the technology is overblown or defective – or both – but is, nevertheless, changing everything for the worse, as corporations replace humans with automated bullshit machines.

These polar opposites have two things in common. The first is that both deploy what Casey Mock of Duke University in North Carolina likens to the Mean Girls playbook – the threat of exclusion and the contempt directed at outsiders.

So for AI boosters, those who don’t buy into the vision “just don’t get it”, while for the doomsters, any indication of a healthy interest in the technology shows that you’re already well on your way to the dark side.

The second shared feature of the cults is a degree of certainty that borders on religious belief; by and large, these are folk who don’t do epistemic humility. Which is a pity, because that’s exactly what we need to make sense of the AI conundrum.

Watching the investment bubble in AI in which companies are planning to spend more than $500bn building AI infrastructure, and struggling to identify any business model that could conceivably service that investment, makes you reach for William Goldman’s 1983 book Adventures in the Screen Trade.

In it, the Hollywood screenwriter, after examining how films get made and why so many predictions fail, concluded: “Nobody knows anything” – which is to say, no one in the entire motion picture industry is certain about what’s going to work.

Which is exactly where we are with AI. Lots of people are trying to figure out how this story will play out in the long term. One source of inspiration stems from Carlota Perez’s 200-year-plus history of technological revolutions, from canals, railways, electricity and automobiles to the computing era. To some analysts, this suggests that what we are living through is not the beginning of a new exponential cycle but just the maturing of the technological revolution that began in the 1970s with the microprocessor and the internet.

From that perspective, the mad rush we’re seeing is merely the biggest beneficiaries of that revolution jumping on AI as the last chance to squeeze more juice from it before it’s exhausted.

The alternative view rejects the “maturity” hypothesis and tries to guess where on an exponential curve today’s AI is located. Recently, Benedict Evans, a shrewd tech analyst, said that AI is at the stage where the world wide web was in 1995 – just when Amazon was getting going and before Google and social media appeared.

It’s a vivid analogy, but it doesn’t hold. The web was simply a platform built on the underlying architecture of the internet, whereas AI looks like being a general purpose technology (GPT) – a core technology that can be used across many industries, improves over time, and stimulates further innovations and productivity gains in a wide range of applications. Like electricity. Or the internal combustion engine. That means it will be everywhere and will affect the world in ways that are hard to imagine.

In that context, the internal combustion engine is an instructive case study. As some American wag observed in 1958, the car “changed our dress, manners, social customs, vacation habits, the shape of our cities, consumer purchasing patterns, common tastes, and positions in intercourse” – but that was all.

The first car powered by a combustion engine was made by Gottlieb Daimler and Wilhelm Maybach in Stuttgart in 1889 but nobody in 1900 could have imagined how the technology would reshape the world – and, in particular, how it would transform cities, laws, industry and employment and social behaviour.

In Britain, for example, the car changed the relationship between the respectable classes and the police. Until the 1950s, private motorists were predominantly middle and upper-class people who had tended to view the police as their servants.

When traffic regulations arrived, though, these same people suddenly discovered they were offenders. Many drivers found themselves confronted with the law for the first time; and, to make matters worse, those enforcing the regulations were police officers, who were generally working class.

GPTs have always changed the world, in other words, and AI will be no different. And if asked where we are on that transition, my guess is somewhere between Daimler’s workshop and Henry Ford’s production line: the technology exists, but the world it will make remains, for now, unbuilt.

Power point

Predatory Hegemons is a cautionary blogpost by Michael Ignatieff examining “the nemesis of power”.

Saving time 

Paul Graham’s The Brand Age is a remarkable essay on the importance of branding in the near-death experience of elite Swiss watchmakers – and their recovery.

Atomic dreams 

Fifteen Years After Fukushima is economist John Quiggin’s scathing commentary on the fantasies of a nuclear renaissance.

Photograph by Alamy