If robots divide us, they will conquer

FT.com

February 4, 2014 6:57 pm

By Martin Wolf

The rise of intelligent technologies may cost us dear – unless we understand the dangers

With one rub of his lamp, Aladdin could command an intelligent being able to fulfil all desires. His genie was a spirit. But the dream of powerful and intelligent artificial servants has also encompassed physical beings. Now, it is becoming a reality built of silicon, metal and plastic. But is it a dream or a nightmare? Will clever machines prove beneficial? Or will they be Frankenstein monsters?

This is the question raised by The Second Machine Age, a new book by Erik Brynjolfsson and Andrew McAfee of the Massachusetts Institute of Technology. This predicts that we will experience “two of the most amazing events in human history: the creation of true machine intelligence and the connection of all humans via a common digital network, transforming the planet’s economics. Innovators, entrepreneurs, scientists, tinkerers, and many other types of geeks will take advantage of this cornucopia to build technologies that astonish us, delight us, and work for us.”

What distinguishes the second machine age from the first is intelligence. The machines of the first age replaced and multiplied the physical labour of humans and animals. The machines of the second age will replace and multiply our intelligence. The driving force behind this revolution is, argue the authors, the exponential increase in the power (or exponential fall in the cost) of computing. The celebrated example is Moore’s Law, named after Gordon Moore, a founder of Intel. For half a century, the number of transistors on a semiconductor chip has doubled at least every two years. Similar progress has occurred elsewhere. (See chart.)

The authors argue that after half a century of progress we are seeing leaps in machine intelligence. As computing power grows exponentially, computers are managing tasks deemed beyond reach a few years ago. Soon, they predict, machine intelligence will be everywhere. They offer as a parallel the story of the inventor of chess who asked to be rewarded with one grain of rice on the first square of his board, two on the second, four on the third, and so forth. Manageable on the first half of the board, the reward reaches mountainous proportions on the second. Our reward will grow similarly.

Yet, to paraphrase a celebrated 1987 quip about computers by Robert Solow, a Nobel-laureate MIT economist, we see information technology everywhere except in the productivity statistics. Trends in output per hour in the US are quite mediocre. Indeed, after an encouraging surge in the 1990s and early 2000s, growth has subsided again. Recent performance in other big high-income economies is worse.

One possible explanation is that the impact of these technologies is overhyped. Not surprisingly, the authors disagree. Indeed, they argue that far from being exhausted, the possibilities are boundless: “digitisation makes available massive bodies of data relevant to almost any situation, and this information can be infinitely reproduced and reused”.

If so, why are measured increases in output so modest? The answers offered are: the plethora of cheap or free services (Skype or Wikipedia); the scale of do-it-yourself entertainment (Facebook); and the failure to account fully for all the new products or services. Before June 2007, an iPhone was out of the reach of even the richest man on earth. Its price was infinite. The fall from an infinite to a definite price is not reflected in the price indices. Again, the “consumer surplus” in digital products and services – the difference between the price and the value to consumers – is often huge. Finally, measures of gross domestic product also underestimate investment in intangible assets.

It seems quite plausible that the proliferation of new gadgets, and the rise of the digital economy with its uniquely low marginal costs, have had a far bigger effect on welfare and even GDP than current measures indicate.

Yet worries remain. The information age has coincided with – and must, to some extent, have caused – adverse economic trends: stagnation of median real incomes; rising inequality of labour income and of the distribution of income between labour and capital; and growing long-term unemployment.

Among the explanations are: fast-rising productivity in manufacturing; skills-biased technical change; the rise of global winners-take-all markets; and the role of rental income, particularly from intellectual property. Think of the difference between the cost of developing Google’s search algorithm and its value. Globalisation and financial liberalisation are also at work, both also boosted by new technologies.

Above all, insists the book, this is just the beginning. Much routine brain-work will be computerised, as happened to clerical skills. Middle-income jobs could hollow out far further. The outcome could be still more polarised incomes, with a tiny group of winners at the top and a vastly larger group struggling below. In 2012, for example, the top 1 per cent of Americans earned 22 per cent of all incomes, more than double their share in the 1980s.

There are good reasons why people should be disturbed by this. First, the lives of those at the bottom might get worse: the authors note that the life expectancy of an American white woman without a high school diploma fell five years between 1990 and 2008. Second, if income becomes too unequal, opportunities for young people dwindle. Third, the wealthy become indifferent to the fate of the rest. Finally, a vast inequality of power emerges, making a mockery of the ideal of democratic citizenship.

In the distant future, thinking machines might even overwhelm our sense of ourselves, just as the best human chess players now know they are not the best on earth. But well before that, the authors suggest that income inequality is likely to increase further, tarnishing the silver age of opportunity the book also promises.

Big challenges arise, then, both now and in the future, if we are to ensure the new machines do not become our Frankenstein monsters. These have big implications for public policy on property rights, education, taxation and other government measures that aim to promote human welfare. I will look at these controversial issues next week.