Nicholas Carr: 'Are we becoming too reliant on computers?'

http://www.theguardian.com/books/2015/jan/17/nicholas-carr-are-we-becoming-too-reliant-on-computers

Version 0 of 1.

Are humans necessary? The answer to that question would seem to be in doubt. As computers continue their inexorable advance, software is taking over sophisticated jobs in fields as various as accounting and medicine, architecture and law. Soon, our Silicon Valley magicians tell us, robots will be driving our cars and flying our planes, diagnosing our ailments and prescribing treatments, teaching our kids and caring for our ageing parents, and, in general, orchestrating our lives. By the end of the next decade, Google’s Ray Kurzweil predicts, artificial intelligence will outstrip the human variety. We’ll have been made redundant by our machines.

It’s true that our jobs and lives are becoming ever more automated, and it seems a sure bet that the trend will continue. Computers aren’t about to slow down or dumb down. But we shouldn’t write ourselves out of the plot just yet. As digital technology sprints forward, we’re not just learning about the possibilities of computer intelligence, we’re also getting a lesson in its limits. The most subtle of our human skills – our common sense, our ingenuity and adaptability, the fluidity of our thinking – remain well beyond the reach of programmers. Far from displacing us, computer automation is underscoring just how valuable, and singular, our own talents are.

Consider the much-celebrated self-driving car. When Google announced, in 2010, that it had built a Prius that could drive itself through traffic – an amazing achievement, to be sure – the company also predicted that fully autonomous vehicles would be on sale by the end of the decade. That’s not going to happen. Lost in the hype about robotic cars has been the fact that human drivers continue to play an essential role in their operation. Robots still get flummoxed by all sorts of odd and unexpected events – detours, leaf-covered roads, gesturing traffic police, wind-blown bits of debris – that human drivers interpret and respond to with remarkable aplomb.

Even if engineers succeed in making a robotic vehicle able to handle 99% of driving situations, robotics experts note, that’s still a very long way from full autonomy. It’s possible to imagine self-driving cars operating independently in tightly controlled circumstances, such as on dedicated highway lanes, but as long as cars have to handle the vagaries of real-world traffic in cities and neighbourhoods, a watchful, adept human will continue to have a place in the driver’s seat.

The same is true in aviation. Autopilots handle the bulk of flying these days, but the technology is far from flawless. When automated systems break down or confront situations their software can’t handle, a pilot has to take over. We learned a stark lesson about the limits of flight automation in 2009 when a US Airways jet lost both its engines after hitting a flock of geese on takeoff from LaGuardia airport in New York. Reacting calmly and brilliantly, the pilot, Chesley Sullenberger, landed the plane safely on the Hudson river. Sullenberger’s feat may have been particularly dramatic, but skilled pilots guide planes out of hazardous situations every day.

The shortcomings of robotic drivers and pilots reveal that the skills we humans take for granted – our ability to make sense of an unpredictable world and navigate our way through its complexities – are ones that computers can replicate only imperfectly. When the going gets tough, the software hands the wheel to the human. With software algorithms now taking on knowledge work, we’re coming to find that our most subtle intellectual skills also remain beyond the reach of programmers. The ability of computers and “big data” algorithms to analyse and make judgments about complex phenomena has come a long way in recent years. What it hasn’t done is obviate the need for human insight and intuition.

In medicine, extravagant promises have been made about how information technology will improve care and cut costs. But recent studies of digital record-keeping and diagnostic systems reveal that they actually tend to inflate healthcare costs, while making no measurable difference in the quality of patient outcomes. Computers can provide valuable information to doctors, but they can’t match an experienced doctor’s ability to grasp the intricacies of a patient’s condition or guide the patient back to health. As the surgeon and author Atul Gawande has pointed out, we can only go so far in routinising medicine. The “expert audacity” of the experienced clinician can’t be replaced by software templates and predictive algorithms.

In education, computers are also falling short of expectations. Just a couple of years ago, everyone thought that massive open online courses – Moocs – would revolutionise universities. Classrooms and teachers seemed horribly outdated when compared to the precision and efficiency of computerised lessons. And yet Moocs have largely been a flop. We seem to have underestimated the intangible benefits of bringing students together with a real teacher in a real place. Inspiration and learning don’t flow so well through fibre-optic cables.

The technophile reader will at this point be pulling her hair out. Sure, she’ll argue, computers have limitations today, but the weaknesses are only temporary. Since the speed of computers continues to advance exponentially, it is just a matter of time before machines will be able to do everything we can do. That’s an understandable response, and a typical one, but it springs from flawed reasoning. What ultimately constrains computers’ ability to repicate human thought has little to do with technical characteristics such as processor speed or memory capacity. It has everything to do with the machines’ lack of being. Even the most accomplished computer systems, University of Toronto computer scientist Hector Levesque writes, remain “idiot-savants”. They can be programmed to perform exceptionally well in carrying out tightly circumscribed mental exercises, such as playing chess or keeping a car centred in a lane, but they “are completely hopeless outside their area of expertise”. They remain prisoners of their coding. Their precision may be remarkable, but it is inextricably linked to the narrowness of their perception.

What makes us smart is not our ability to run lots of data through the circuits of our brains or to tackle well-defined problems through scripted routines. It is our ability to make sense of things – to weave the knowledge we draw from observation and experience into a rich and fluid understanding of the world that we can then apply to any task or challenge. It’s this mental suppleness, born of having an awareness of ourselves and our surroundings, that allows us to think conceptually, critically, metaphorically, imaginatively. The great advantage we have over our computers, in short, is that we are alive and they are not. Science-fiction fantasies aside, we are not going to lose that advantage anytime soon.

The real danger we face from computer automation is dependency. Our inclination to assume that computers provide a sufficient substitute for our own intelligence has made us all too eager to hand important work over to software and accept a subservient role for ourselves. In designing automated systems, engineers and programmers also tend to put the interests of technology ahead of the interests of people. They transfer as much work as possible to the software, leaving us humans with passive and routine tasks, such as entering data and monitoring readouts. Recent studies of the effects of automation on work reveal how easily even very skilled people can develop a deadening reliance on computers. Trusting the software to handle any challenges that may arise, the workers fall victim to a phenomenon called “automation complacency”. Their attention drifts, they fail to exercise their skills and their talents begin to wither. In the US, the Federal Aviation Administration has found evidence of a loss of manual flying ability among pilots, which in the worst cases has led to accidents. The agency is now encouraging commercial aviators to turn off the autopilot during flights and do more manual flying. Studies of accountants, radiologists and other professionals have also documented signs of “deskilling” as a result of automation.

If we design and use our thinking machines wisely, they will open new possibilities for us. They will help us solve hard problems, and they will push us to greater achievements. If we stay on the current path, we will doom ourselves to less interesting work and a steady erosion of skills. The suggestion that algorithms are smarter than we are will become self-fulfilling. Certainly, we should respect the capabilities of our computers. But we should respect our own talents even more.

Nicholas Carr’s The Glass Cage: Where Automation Is Taking Us is published by Bodley Head.