A seasoned columnist and accomplished interviewer in his role as Fortune magazine's editor-at-large, Geoff Colvin, is not the only member of the fourth estate to wince at the notion of computers writing stories, which is now a reality in US journalism.
Last year, the Associated Press agency, for example, assigned computers to write all of its articles about corporate earnings announcements. Other media companies, including Yahoo and Forbes, also use articles prepared by a company called Narrative Science.
The machines are getting better at their task, it seems, and can now do endless research to provide context. They can even spot angles, such as the drama of heroic comeback wins in sports fixtures.
Writers who “never get blocked, tired or drunk” have their place, Colvin admits, while making the case at the same time for the primacy of original, human-generated journalism, pieces that capture the quirky throwaway remarks at the start and end of an interview or that include shrewd observations of a subject’s body language, for example.
Whether anyone will be able to tell the difference soon is a different matter. One of the key observations from Colvin’s new book is that underestimating the power of computers is a bad move. Asking what computers can’t do isn’t a dumb question, but it misses the point, he says.
Hubert Dreyfus, of Massachusetts Institute of Technology (MIT), in a book whose title posed that very question in 1972, suggested that computers couldn't make significant further progress in playing chess than the mediocre level then possible. Yet a computer beat world champion Garry Kasparov in 1997.
Similar predictions were made about driverless cars as recently as 2004 in Frank Levy and Richard J Murnane's book The New Division of Labor: How Computers Are Creating the Next Job Market. This suggested that driving a vehicle required such a mass of sensory inputs and complex split-second judgments that it would be difficult for a computer to ever handle the job. Within six years, however, Google had introduced its first autonomous car.
“The pattern is clear,” Colvin notes. “Over and over again, smart people note the overwhelming complexity of certain tasks that humans do, and conclude that computers won’t be able to master them, yet it’s only a matter of time before they do.”
Technology is different
Most things stop growing or at least advancing at pace when they get to a certain stage, but the message here is that technology is different. Moore’s Law, about the doubling of the power of technology every two years, posits a future only imaginable in science fiction, if viewed through the lens of today’s norms. In a straight battle against humans, the machines will win every time.
As Colvin sees it, it isn’t even a simple question of hard versus soft skills. Computers can be programmed, for example, to react to the hundreds of facial clues humans display in everyday interactions. A machine can outdo us even in detecting human emotions, it seems.
Moreover, it can even intervene when our erratic behaviour becomes dangerous to others. For example, a prototype stress-monitoring car, tested at MIT, senses when a driver is highly stressed and is able to suggest relaxing music and give the GPS guidance a more soothing tone of voice.
Researchers have even experimented with special thermochromatic paint to change a car’s external colour based on the driver’s emotional state, as a warning signal to other drivers.
The man-versus-machine issue, however, is not so much about ability as it is about choice.
“There are certain tasks that we want humans to remain in control of, leadership and judgment roles being amongst them,” says Colvin. “It’s not that we are necessarily better in these roles, it’s just that society chooses that these are areas that we should retain control of.”
A classic example is the judiciary. Colvin has an enlightening passage in his book about research into the effects of lunch breaks on judges hearing parole cases. In the resulting study, about 35 per cent of parole applications were ultimately approved, but the approval rating shrunk steadily in the run-up to lunch each day, picking up again in the early afternoon before declining again.
As Colvin notes, it’s almost certain that, given the appropriate data, computer analysis, in the majority of cases, could judge parole applications more effectively and less capriciously than humans could. But the question is: would people be happy with this?
Given a choice, seriously ill patients will want their diagnosis and prognosis delivered by a human. There’s a case that the outcome of their illness will be improved by this, as they are more likely to follow a treatment plan delivered by a doctor than one outlined by a machine.
Generally, patients also want their institutions to be controlled by human beings, albeit with increasing input from computers that can ensure better decision-making.
Left-brain skills
According to Colvin, the technical, left-brain skills that have been in such demand since the Industrial Revolution are no longer the most important ones. Rather, the most human of skills, such as empathy, teamwork and relationship building, will be the ones most valued in the years ahead.
“We’re right to focus on the need for greater education in the Stem subjects, but that focus needs to be tweaked,” he says. “Retailers are telling us that they don’t want programmers now who just sit coding in their own silos; they want people who have the skills to interact with customers in the real world.”
The message here is ultimately a positive one for humanity.
This is not so much a passionate polemic as an extended piece of journalism, with Colvin showing a magpie-like tendency to pick up juicy nuggets and anecdotes and to weave them into an entertaining narrative.
How long it is before the likes of Colvin are replaced by a machine will be interesting to watch.
Humans Are Underrated: What High Achievers Know That Brilliant Machines Never Will, by Geoff Colvin, is published by Nicolas Brealey Publishing
