Technology Shall Lead Us

machine

Erik Brynjolfsson and Andrew McAfee’s The Second Machine Age attempts to predict the pace of change in computer intelligence and weigh its economic implications. They argue that progress is accelerating and that we should be thinking about the opportunities and challenges this poses for society.

Digital technologies “started surprising us,” say the authors, which is why they wrote the book:

Computers started diagnosing diseases, listening and speaking to us, and writing high-quality prose, while robots started scurrying around warehouses and driving cars with minimal or no guidance. Digital technologies had been laughably bad at a lot of these things for a long time—then they suddenly got very good.

Their assessment that rapid progress is taking place is open to dispute. Speech-recognition is still highly erratic. For my high school statistics and economics students, I have lectures on YouTube. The students discovered that by turning on close-captioning, they can see an automated speech-to-text conversion. Hilarity results.

I have been following the topic of digital intelligence since reading Ray Kurzweil’s 1999 book The Age of Spiritual Machines. Kurzweil updated and expanded his thesis in 2006 in The Singularity Is Near. Kurzweil emphasizes the importance of Moore’s Law, which is an observation about the rate at which the price-performance characteristics of computer chips improve as engineers discover ways to improve on their design and manufacture.

Roughly speaking, the power of computer chips doubles about every two years. This exponential improvement is the critical driver of the progress forecast by Brynjolfsson and McAfee as well as by Kurzweil.

Kurzweil believes that Moore’s Law puts computers on track to catch up with and then rapidly surpass the human brain in terms of intelligence. At longbets.org, a web site conceived by Stewart Brand to stimulate thinking about long-term issues, the very first bet placed was by Kurzweil, who wagered $20,000 against Mitch Kapor that by 2029 a computer will be able to pass the Turing Test, meaning that its conversational abilities will be indistinguishable from those of a human being.

The Age of Spiritual Machines had a chapter called “2009” in which Kurzweil made a number of predictions, most of which have proven overly optimistic. He wrote that by that year, most communication with computers would be via speech. Also that computer translation between languages would be routine. He thought people would have perhaps a dozen computers embedded in clothing and jewelry. He described a business world in 2009 in which at least half of all transactions would be conducted on line. “Intelligent assistants which combine continuous speech recognition, natural-language understanding, problem solving, and animated personalities,” he wrote, would routinely help us find information and conduct transactions.

By then, Kurzweil thought, computers would be the main delivery vehicle for content in education, with teachers “viewed more as mentors and counselors than as sources of learning and knowledge.” He predicted that medical diagnosis would almost always involve collaboration between a human physician and an expert system.

All of this was supposed to have been achieved five years ago. Computer chips have attained the power predicted by Moore’s Law, but computer applications have not attained the capability forecast by Kurzweil.

Back at the turn of the millennium, these applications seemed to Kurzweil to be on the near-term horizon. These strike me as the same applications that Brynjolfsson and McAfee suggest are on the near-term horizon today. While a few of Kurzweil’s other predictions did materialize, and while some of these applications are certainly closer to reality today than they were in 1999 or 2009, we should be wary that some of what The Second Machine Age tells us to expect may not in fact appear for several decades, if ever.

If, on the other hand, Brynjolfsson and McAfee are correct, then we are at an inflection point at which capabilities that a few years ago seemed hopelessly out of reach for computers are now approaching fruition. They cite the progress with Google’s driverless car, IBM’s championship-caliber program to play Jeopardy, and the development of robots that are more nimble and trainable than were previous iterations.

In addition to Moore’s Law, the authors cite another factor that makes them optimistic. We now have many more people around the world able to communicate with one another. They see this as likely to increase the rate of innovation. In poor countries, the rapid acquisition of cell phones has suddenly given billions of people access to Internet resources, which raises their potential to acquire knowledge and contribute to science and technology.

Brynjolfsson and McAfee point out that some of the benefits of information technology may be understated by GDP statistics, which measure output in terms of transaction prices. The benefits to consumers of goods and services can in fact be much more than what we pay for them. This is particularly striking in the case of the Internet, where so much information and entertainment is available at no charge whatsoever.

Another result of digital technology is what the authors call “spread,” or what other economists have called “winner-take-all” or “winner-take-most” markets. In mass-market photography, they contrast an analog-age success story, Kodak, which at one point had nearly 150,000 employees, with a digital-age success story, Instagram, which was created by a company of just 15 people. After this book came out, Facebook’s acquisition of WhatsApp, a 30-person company, for $19 billion serves as yet another example.

The contemporary market structure leads to much wider dispersion in the distribution of wealth because, as Brynjolfsson and McAfee write:

More wealth will be created by less work . . . If the work a person produces in one hour can instead be produced by a machine for one dollar, then . . ..either that worker must accept a wage of one dollar an hour or find some new way to make a living. Conversely, if a person finds a new way to leverage insights, talents, or skills across one million new customers using digital technologies, then he or she might earn one million times as much as would be possible otherwise.

Some of what appears to be cyclical unemployment may in fact be the unfolding of the process of workers being displaced by information technology. The authors describe a candid discussion they had a few years ago with a CEO:

He explained that he knew for over a decade that advances in information technology had rendered many routine information-processing jobs superfluous. At the same time, when profits and revenues are on the rise, it can be hard to eliminate jobs. When the recession came, business as usual obviously was not sustainable, which made it easier to implement a round of painful streamlining and layoffs. As the recession ended and profits and demand returned, the jobs doing routine work were not restored. . . . his organization found it could use technology to scale up without those workers.

Technology thus provides a mixed blessing for low-skilled workers. Many goods and services are less expensive, but prospects for wages and employment have dimmed.

Brynjolfsson and McAfee present both sides of the debate over whether living standards have improved in recent decades for those near or below the middle of the wage distribution. On the one hand, looking at the cost of food and consumer durables, median real wages have increased substantially. On the other hand, if we include the cost of health insurance and higher education in the market basket, median real wages have stagnated.

Despite their concerns that innovation is a mixed blessing, the authors would like to see it encouraged. They recommend, among other policy ideas, reform the U.S. intellectual property regime, particularly when it comes to software patents and copyright duration . . . [patents] have to strike a delicate balance; they have to provide enough intellectual property protection to encourage innovation but not so much as to stifle it. Many of today’s informed observers conclude that software patents are providing too much protection. The same is probably true for at least some copyrights.

A patent is in effect a monopoly granted and enforced by the government for a period of years. It would be preferable not to grant monopolies at all, if innovation can be encouraged in other ways. For example, Brynjolfsson and McAfee discuss the use of prizes for innovation. They point out that prizes work well if a problem can be clearly defined so that one can easily determine if a proposed solution merits reward.

Overall, The Second Machine Age could be useful to those looking for a current perspective on the economic issues raised by exponential improvement in information technology. Bear in mind, though, that anyone who has read Tyler Cowen’s Average Is Over, or the book these very authors turned out in 2012, Race Against the Machine, will find the analysis here, and even some of the examples used, a little familiar.

Arnold Kling

Arnold Kling was an economist on the staff of the Board of Governors of the Federal Reserve System from 1980-1986. He was a senior economist at Freddie Mac from 1986-1994. In 1994, he started Homefair.com, one of the first commercial sites on the World Wide Web. (Homefair was sold in 1999 to Homestore.com.) Kling is an adjunct scholar with the Cato Institute and a member of the Financial Markets Working Group at the Mercatus Center at George Mason University. He teaches statistics and economics at the Berman Hebrew Academy in Rockville, Maryland.

About the Author

Comments

  1. says

    In my opinion there will never be a singularity. Computer intelligence will not increase exponentially. Machines will be better at individual cognitive tasks than humans, but machine “intelligence” will only approach the power of the human mind asymptotically. The reason for this is that, in order to adapt and accommodate novelty, a mind, human, machine or otherwise must be prone to error. Deterministic processors, i.e. machines, become more efficient by reducing error. This makes them “smarter” in some ways but the more tolerance of error is reduced, the less able they are to accommodate ambiguity. The smarter they become the less flexible they are. Humans tolerate both novelty and ambiguity with neat trick: human thought processes become both more efficient and less efficient at the same time. When this paradoxical power is lost or impaired, mental illness results, even in people of astonishing intelligence. Idiot savants have a type of genius, like most geniuses, for which highly developed abilities are not dependent on experience; yet in other ways they are cognitively impaired.

    It is not the potency of individual cognitive functions that makes human intelligence remarkable; it is the balance between creative and destructive, deterministic and stochastic, appealing and aversive, and the way that these change continuously that machines can only approximate.

Trackbacks

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>