Dan O'Brien: The hopes and fears around artificial intelligence
Published 14/08/2016 | 02:30
'Artificial intelligence could spell the end of the human race." That is how no less a figure than physicist Stephen Hawking has described the risks posed by intelligent machines capable of thought and action.
He is not alone. Some, if by no means all, of the world's deepest thinkers in relevant fields are profoundly worried about the possible negative consequences of rapid and accelerating advances in artificial intelligence (AI).
The happening that triggered the latest wave of concern took place in March of this year. Then, a milestone was reached when the computer programme, AlphaGo, defeated a human in a game of Go - a Chinese board game previously been considered too complex for a machine ever to best a human.
AI advances are, as always with new technologies, being driven by those who see more opportunities than threats in change. As the opportunities are almost limitless, companies are now pouring massive sums into AI in the hope of making the next breakthrough. And of reaping the returns.
Machines already do much of the work once done by people's muscle power. The next step is to complete more tasks requiring brain power.
There is a distinction between 'narrow AI', which can perform specific tasks, and stronger forms of AI, which would perform a full range of human cognitive abilities. The narrow form has been around for years, and nearly everyone has encountered it. Examples includes AI that can filter spam emails and give purchase recommendations on websites. As we all know, these forms of AI can be pretty dumb.
But better stronger forms of AI are drawing closer, even if estimates among tech types on when they will be widely available vary a great deal. How advances in AI and other technologies will affect different aspects of our lives divides opinion too.
At one extreme are the utopians who envision a world in which our lives improve immeasurably. At the other are dystopians who see multiple negative consequences, including machines replacing humans in an increasing number of jobs and AI being used against humanity, either by other humans or (as in the Terminator movies), by machines themselves.
As AI advances in the coming years, a number of important economic, ethical and practical concerns will inevitably come to the fore. For instance, should machines be programmed to mimic humans or be distinct? Who will controls AI once it is developed - the developers exclusively, or will new forms of regulation have to be created? And could AI create unprecedented levels of inequality as the gains go to the small group who own the technology while the losses are counted in mass unemployment.
For a column focusing on economics for a business readership, the most relevant issues relate to the impact of AI on companies and the people who work in them.
It is perhaps fair to say that debate thus far on the topic has focused more on the downsides than the upsides. The idea that automation is causing joblessness and inequality is prevalent and oft repeated. But while the likes of Hawking are right to highlight the risks of AI in the longer term, I am on the more optimistic side of debate when it comes to employment issues.
Partly explaining the pessimism is fear of change and the unknown. Another reason is what economists call the "lump of labour fallacy" - the idea that the number of jobs in an economy is fixed and, therefore, that more automation must mean fewer jobs.
Even the briefest glance at history shows that this notion is not called a fallacy for nothing. Going back as far as Ireland's first export industry - stone-age axes - technology has never stopped destroying jobs, but new jobs have always emerged to replace them. Today, there has never been more people at work on the planet.
Even in the rich world, where fear of losing out to both technology and low wage economies, nearly every economy has got more of its people working than 20 years ago. As the chart shows, the US has been (almost) exceptional, which may well explain in part recent political happenings.
For starters, jobs will be created directly to service any new technology. Moreover, greater automation tends to boost productivity, the engine of increased prosperity. That, in turn, allows people to consume and invest in other areas instead. Labour is reallocated, rather than displaced entirely.
Martin Ford, author of The Rise of the Robots: Technology and the Threat of Mass Unemployment, believes this time is different. He argues that because almost every sector uses computers, there will be no safe havens when it comes to job displacement. Given the breadth and speed of coming change, it will be more difficult for workers to move to other jobs than the past.
Some economists are also worried future economies will consist of low-skilled jobs that are not profitable to automate and high-skilled jobs that complement new technologies. The losers will be a widening swathe of middle-skilled workers.
Because the trajectory of future technology is unpredictable, the exact impact on jobs is rather unclear. A much cited study by researchers at Oxford University found that 47pc of jobs in the US could become automated in the next 20 years. Yet a recent OECD paper* found (more plausibly, in this columnist's view) that 9pc of jobs at highrisk of being automated. The figure for Ireland was even lower, at 8pc.
If opinions differ greatly on how negative the effects might be, there is more consensus on the sectors most likely to be disrupted by smarter machines. Those working in office support, retail and transport are seen as being affected most. Those in creative jobs, by contrast, are least likely (in the medium-term at any rate) to be replaced by machine.
What most experts agree on is the need for governments to act in order to smooth out the disruption technology is expected to bring. After all, industrialisation eventually forced states to introduce labour regulations and measures that became the welfare state. It seems most likely that education systems offer the best hope of ensuring as many people as possible access new opportunities and as few as possible are, to use a voguish phrase, left behind.
Among other things, lifelong learning and re-skilling are likely to be a feature of working lives from now on. Innovation in education can help this process. There has been an emergence of Massive Open Online Courses (MOOCs) where people can take online classes on platforms such as Coursera, edX and Udacity (incidentally all three were founded by people involved in AI).
When these websites were launched a few years ago, the hype that they might displace universities was overblown. But they will provide at least a part of people's formation in the future.
The prospect of people constantly updating their skills and working a number of jobs in their lifetimes seems possible, particularly for those with high-level and technical skills. More focus will be required for those with low- and middle-level skills.
AI technology brings with it challenges and opportunities. It will doubtless bring great change to the economy and society.
The important lesson is that when faced with change, the best course of action is usually to adapt proactively.
Sunday Indo Business