My recent book, The Economic Singularity, argues that improvements in machine intelligence are going to make it impossible for many of us –maybe all of us – to earn a living. If this is even an outside possibility, we’d all be wise to spend some time working out how we intend to deal with it.
In 1900, 40 per cent of American workers were employed in agriculture. That has now fallen below three per cent and while the farm workers eventually found better jobs elsewhere in the economy, their horses didn’t. For the year 1900 was ‘peak horse’, with around 25m of them working on farms in America. Today, there are virtually none. The difference between the fate of human and horse farm workers is that once machines provided the muscle, humans had something else to offer: cognitive, emotional and social abilities. Since horses had nothing else to offer, their population collapsed. The question now is whether we are approaching the ‘peak human’ era.
Artificial intelligence is improving at an exponential rate but machines are still far from becoming Artificial General Intelligences (AGIs) –possessing the same flexible cognition of an adult. Even the smartest machines possess only narrow intelligences: they may play a mean game of Go, but they can’t experience fear and they can’t yet write a good joke. Most significantly, even though they can beat a human opponent at Go, they aren’t even aware they are playing a game.
But machines don’t need to become AGIs to displace most of us from our jobs – they simply have to become better than us. And because machines excel at many forms of pattern recognition, including natural language processing and image recognition, they’re already well on their way to overtaking us. And, of course, once a machine has your job, it will do it faster, better, longer and cheaper than you can, with its abilities improving exponentially all the time.
“To society at large, self-driving cars will be the canary in the mine – the single development that serves as a wake up call.”
But the fate of drivers is likely to be swiftly shared by other sectors. Some think that mass unemployment fears are premature since society will develop new human roles as machines take over the old ones. Even if this is possible, these are only likely to be temporary respites as whatever these new jobs are, machines will be more capable of doing them too. In the medium term, it does seem likely many will not be able to earn a living.
While that sounds terrible, it could actually be extremely good news. A world in which machines do all the boring work could be wonderful. They could be so efficient that goods and services could be plentiful and, in many cases, free. Humans could get on with the important business of playing, relaxing, socialising, learning and exploring. Surely this is what we should be aiming for?
Of course there are challenges, and my book explores these in detail. With considerable reluctance, I conclude that we may need to adopt an entirely new economic system. We don’t know what this should look like, or how to get there, which is why I talk about an “economic singularity”, borrowing the term from physics to describe a point at which the normal rules cease to apply.
The book argues that we should be monitoring our economies for signs that technological unemployment is beginning, and undergoing careful scenario planning to help us to
avoid the negative outcomes and ensure the good ones. As we face the prospect of widespread technological unemployment, this planning may turn out to be mission critical for our economies, and our civilisation.
Calum Chace is an author and speaker about artificial intelligence’s likely future impact on society and will be speaking at Disruption summit Europe.