Doom or boom — what is our AI future?

Excited and worried by the impact of AI, Rishi Sunak arranged — to much ballyhoo — his AI Safety Summit, which will begin on Wednesday in Bletchley Park, Milton Keynes. So which is it: doom or boom? Are we entering a new golden age of humanity? Or the last age of humanity?

Tyler Cowen, an economist and author, is more excited than panicked at the possibilities of AI

RAY WELLS FOR THE SUNDAY TIMES

One person who has no time for the “doomers” is Tyler Cowen. Once dubbed America’s “hottest economist” by Time magazine, Cowen, the author of more than 20 books and the hugely popular Marginal Revolution blog, is helping to set the terms of the debate on how AI may transform our world.

In a speech last week, the prime minister echoed Musk by issuing his own “stark warning” of the dangers ahead. AI could be used to build bioweapons, flood the airwaves with misinformation, sabotage elections and supercharge child sexual exploitation, Sunak said. We might “lose control” of the machines altogether.

Cowen views such catastrophism as a waste of time. “I’ve issued the following challenge to the doomers: be like the climate change people. Publish your models. Do it in the arena of science. I will listen. Make your case.” He adds: “There’s not one published model of [out-of-control AIs] in peer-reviewed journals. The doomers are losing the argument.”

Cowen, 61, sides with the boomers. He’s more excited than panicked. “The closest historical analogue to AI is the printing press, which also spread a kind of intelligence,” Cowen says. The printing press indirectly led to a catalogue of horrors, from religious wars to Chairman Mao’s Little Red Book. “But we would never press the ‘No’ button on the printing press,” he adds. “We’re seeing a sped-up version of those effects, in a way that will be wonderful, and will make most people much better off.

“For the first time, humanity has created a kind of genuine intelligence. For matters intellectual or visual, it can perform remarkably well. It can do better than humans at many tasks within a fixed environment. And now we have it. It’s a miracle and a blessing that the freer nations have the quality version first, and I think we now need to decide what we’re going to do.”

In some realms, the revolution has already begun. Software developers who use GitHub’s “co-pilot”, an AI code-writing assistant, reported a 55 per cent decrease in the time it takes to complete a task. This is tantamount to gaining a superpower, virtually overnight.

Vast amounts of software are now entirely self-generated by AI. Emad Mostaque, chief executive of the London-based developer Stability AI, predicted that the jobs for most of India’s outsourced software developers — the country is home to five million coders — will be “gone” within two years. Other jobs are not long for this brave new world. The bottom and soft middle of the white-collar pyramid — call centre staff, customer support, office workers doing rote or repetitive tasks — will be the first to go, Cowen reckons. But gardeners, carpenters and other “real world” specialists, where automation cannot whittle down the price of their labour to zero, should fare well, he argues.

Human creativity and dexterity can sometimes be underrated — in the new biography of Musk, he finds that humans are actually more efficient at assembling some of the cars than robots, and so fires the robots.

Elon Musk is said to prefer work done by people over robots but Tesla is developing fully self-driving cars and even a humanoid robot “Optimus”

CHRISTIAN MARQUARDT/GETTY IMAGES

Not everyone shares this measured view. Vinod Khosla, the legendary tech investor, recently predicted that within 20 years there will be one billion AI-powered bipedal (two-legged) robots in the world. “I do think we will, within some period of time, have the ability to make work optional,” he said. “We’re getting to the point where it will be interesting to ask what it means to be human.” Indeed, even Musk, perhaps the world’s most prominent AI alarmist, has predicted that the bipedal robot being developed by his company Tesla will be bigger than its $650 billion car business.

Such excitable predictions are, of course, typical of Silicon Valley. And they fuel the doomerism that has taken hold in government. Cowen, however, is not convinced that it will play out in the way that the boomers — or the doomers — expect.

Things may change more slowly than we think. If the history of technology has taught us anything it is that the power of inertia is immense. Uber, for example, promised to replace car ownership with “on-tap” transport as reliable as running water. We have been told for years that radiologists would be replaced by AIs far superior at interpreting x-rays, yet this day seems scarcely closer. “The human bottlenecks are pretty severe,” Cowen says. Remaking systems, institutions and belief systems is painful, slow, and vested interests have creative ways to hold on to power. “Things will take longer than many people realise. I think we’ll stay at full employment for the indefinite future,” Cowen predicts. “The 20-year impact is where it will really matter.”

Make no mistake, however. The years in front of us will be “messy” and “scary”, Cowen argues, because AI is a “skewed accelerant”. Industries will be washed away. New ones will emerge. Income disparities could widen dramatically, opening deep societal fissures. Millions may lose their jobs. But they will also have access to free, or nearly free, personal tutors to help them retrain, as will hundreds of millions of children in the developing world who lack access to high-quality education. Medical researchers are using generative AI to imagine novel medicines such as cancer vaccines.

All of which suggests we are on the cusp of a new era. The last several decades, in which America was the unrivalled global superpower and technological change was fast but not radical, were an aberration, Cowen argues, a bubble “outside of history”. The arrival of AI has dropped us back into “moving history”, a time of tech-infused tumult. “We will have to reconceptualise who we are, what is intelligence, what are we truly good at, what does beauty mean to us?

“Will we be letting AI help raise our kids? How does it intersect with religion? How do I know what job I’ll have? All of this will change,” says Cowen. “And we are not psychologically equipped for it. This, I think, is the greatest danger: the intersection of dynamic AI with our rather static attitudes, status quo biases, inertia and sclerotic institutions.”

Rishi Sunak spoke of the dangers of AI

PETER NICHOLLS/PA

Sunak’s AI summit appears to be a calculated, if slightly desperate, attempt to ensure that Britain does not get left behind. He wants to turn this island into a global centre of excellence by creating an “AI safety institute” that will police the latest AI models all over the world and turn Britain into a magnet for talent serving as a bridge between the economic freedoms of America and the bureaucratic caution of the European Union. The two-day event will be attended by a select group, including the US vice-president Kamala Harris, AI experts, and very controversially, a Chinese delegation. Many other key figures, including President Macron and Olaf Scholz, the German chancellor, will be absent, however.

In America and among the Tory hawks surrounding Liz Truss, the former prime minister, China’s inclusion has stirred unrest. The view from Washington is that the West is locked in a “techno-economic war” with Beijing that is not unlike the Cold War. Seen through that prism, whoever gains superiority will rule the world. “The cat is out of the bag,” says Cowen. “We need to work hard to build more good stuff, before other people build more bad stuff.”

This is why Mark Zuckerberg, who co-founded Facebook and its parent company, Meta, has found himself at odds with the Pentagon. The 39-year-old billionaire has “open-sourced” its Llama 2 large language model, meaning he has made the company’s ChatGPT-like tool free, including access to its code and the ability to build apps on top of it. Khosla argued that doing so was a “dangerous” decision akin to “open-sourcing the Manhattan Project”.

When future anthropologists look back on this era, the central tool, the totem of our times, will be the smartphone. Its days are numbered, says Cowen. The magic of AI is that it has cracked natural language. If the machines have learnt our human operating system, why should we keep tapping tiny rectangular screens when technology simply understands us in a way that it never has before? Might our device of choice morph into, say, a voice assistant accessed through a lapel pin, earpiece or glasses?

Gardeners are probably safe from AI taking their jobs, for now

GETTY IMAGES

OpenAI is said to be considering teaming up with Sir Jony Ive, the iPhone designer, on a reimagined, AI-centred smartphone successor. Meta recently announced AI-enabled glasses that respond to voice prompts and produce information on objects that appear in one’s field of view. “We’re going to have new devices, and I think fairly soon,” Cowen says. “The browser will be displaced.”

Cowen is an early adopter. Last week he published what he claims is the first “generative” AI-written book, Goat: Who is the Greatest Economist of all Time and Why Does it Matter? Cowen wrote the book, but has integrated it with ChatGPT, allowing readers to instruct the AI to turn a chapter into a limerick, generate a quiz or query assertions that Cowen makes.

It will not be long, he reckons, before news stories, podcasts and other media will also be published as “living” documents that can be actively queried and interacted with. The “static” story will soon be dead, as will so much else that we take for granted.

So does Cowen have any suggestions for how we navigate this moment of “moving history”? How do we steer our children toward a life that is not stripped of meaning and value? On that question, sadly, he doesn’t have a reassuring answer: “Try to learn things in small bites. Don’t let it scare or intimidate you. And get in a group of people with common interests who are trying to figure out what’s happening. I think that’s the best I can say to most people.”

One can perhaps take solace in Cowen’s prediction that the day we are rendered useless is still some way off. Unless, of course, you’re a gardener and you don’t have to worry at all. Cowen says: “It will be a messy but wonderful history that I feel privileged to be able to witness.”