Listen to this article
If Daniel Nadler is right, a generation of college graduates with well-paid positions as junior researchers and analysts in the banking industry should be worried about their jobs. Very worried.
Mr Nadler’s start-up, staffed with ex-Google engineers and backed partly by money from Google’s venture capital arm, is trying to put them out of work.
Its algorithms assess how different securities are likely to react after the release of a market-moving piece of information, such as a monthly employment report. That is the kind of work usually done by well-educated junior analysts, who pull data from terminals, fill in spreadsheets and crunch numbers. “There are several hundred thousand people employed in that capacity. We do it with machines,” says Mr Nadler. “We’re not competing with other [tech] providers. We’re competing with people.”
Warren– the name given to the system, in homage to investor Warren Buffett – is part of a new army of “smart” machines that are threatening to invade office life. These computers do not just collect and process information; they draw inferences, answer questions and recommend actions, too.
The threat to jobs stretches beyond the white-collar world. Advances in artificial intelligence (AI) also make possible more versatile robots capable of taking over many types of manual work. “It’s going to decimate jobs at the low end,” predicts Jerry Kaplan, a Silicon Valley entrepreneur who teaches a class about AI at Stanford University. Like others working in the field, he says he is surprised by the speed at which the new technologies are moving out of the research labs.
“People don’t understand it, they don’t get what it’s going to mean,” adds Mr Kaplan, who says he was “radicalised” by a sudden awareness of the job-destroying capabilities of the new technology. “I feel like one of the early guys warning about global warming.”
The impact of IT and automation on the world of work – and dire warnings about the job destruction they might cause – are as old as the technology itself. But the convergence of a number of tech trends has made the threat more immediate.
As a result, 47 per cent of jobs in the US are now at risk from computerisation, according to a prediction last year from Carl Benedikt Frey and Michael Osborne from Oxford university. McKinsey, the management consultancy, has estimated that by 2025, productivity gains in fields of “knowledge work”, ranging from clerical to professional services, could account for 40 per cent of all the current jobs in those areas.
One long-familiar tech trend is the relentless fall in the cost of computing power. According to Erik Brynjolfsson and Andrew McAfee, academics at the Massachusetts Institute of Technology, these incremental advances in computing have combined to make great leaps. Their book, The Second Machine Age, has stirred up angst this year about what the coming smart machines will do to job levels.
A second factor is the availability of vast bodies of digital data. Feeding off that information, advanced pattern-recognition systems – using a technique known as machine learning – are able to draw deductions that earlier machines could not attempt.
This has given rise to a field known as “cognitive computing”. IBM has been the acknowledged leader since its Watson machine won a television quiz in the US three years ago, in the process overcoming the notoriously difficult challenge for a computer of mastering “natural” language.
New ways of interacting with computers, making it easier for non-technical humans to work with machines on complex tasks, are a third part of the AI revolution. As with many other aspects of technology, smartphones have been the forerunner.
Services such as the Siri question-and-answer feature on Apple’s iPhone and the Google Now service that tries to anticipate a user’s information needs, may be the forerunners of similar user-friendly systems that will come to dominate the office.
This has fed two visions of the future of work. In one, the machines take on many of the boring parts of a job, setting humans free to supply the more advanced – and satisfying – brain work. The other vision is less harmonious: the machines leave many human workers on the scrap heap altogether.
The incursions being made by this new generation of technology can be hard to trace, although that does not diminish its potential impact.
A Los Angeles-based company called SmartAction, for instance, uses machine learning and natural language recognition in its automated response software for call centres. The better it can “understand” what callers want, the more likely the technology can deal with their queries successfully and prevent a human call-centre worker getting involved, says Tom Lewis, chief executive.
“It makes it so you need fewer agents,” he says. And while many baby boomers may still yearn for a human voice, he adds, a younger generation brought up on digital technology feels perfectly at ease dealing with automated systems.
In other fields, the displacement of humans can be easier to spot. With technology from a Chicago company called Narrative Science, Len Welter, an entrepreneur in London, used machines to write reports based on financial data. His start-up, which was sold last year to British financial information firm Markit, produced a “newswire” of automated reports to round out deeper research being done by human analysts.
Despite churning out 40 reports a day, he claims the robo-writer, called Quill, was good at disguising its non-human origins: “They changed the grammar and language – you couldn’t tell it was from a computer.” He admits the human writers at his company “freaked out” when they heard he was planning to use the system.
The workers that machines threaten to displace cover a wide range of office work. Smart digital assistants, for instance, could stand in for many types of support staff – or, by making the ones who remain more productive, greatly reduce their numbers.
The jobs of many analysts and researchers could also be in the line of fire. Advances in machine learning and natural language systems make it easier to interrogate large amounts of data and to derive smarter answers in more intelligible forms.
Even highly paid professionals with specialist expertise are not immune. In fields such as law and medicine, machines are likely to produce “generally better answers” than humans, who struggle to keep up with the latest knowledge in their fields, says James Manyika, a director at McKinsey Global Institute.
IBM recently began selling its most advanced cognitive computing technology as an “ingredient” to be used in other business software applications – a kind of “Watson Inside” approach. Moves like this could see deeper intelligence seep into a wide range of everyday office technologies.
Perhaps not surprisingly, most of the technologists working in the field – as well as the companies buying the new systems – predict that the upshot of moves such as this will be to enhance human workers rather than replace them altogether. The new systems are promoted as opportunities to jettison the most dull and onerous aspects of work.
Companies involved in digital advertising, for instance, engage in big number crunching to improve their campaigns, making constant data-driven judgments about which sites to use, at which times of day and in front of which types of audience.
“It’s a real drag to have that job – it’s a constant, crushing load,” says George John, chief executive of Rocket Fuel, a digital advertising company whose technology instead automates the process.
Robotics companies trying to replace various types of manual labour make similar claims for their products, notes Mr Kaplan. A weeding machine smart enough to pick its way through a field without harming the crop, for instance, is advertised by its manufacturer, Blue River, as an advance on the “back-breaking job” it is replacing.
However, for the humans who risk being put out of work, Mr Kaplan adds, it may be little consolation to be told that their jobs were undesirable in the first place.
A second defence of the smart machines is that the slow speed of human labour causes bottlenecks in an increasingly digital production chain. Even an army of analysts with calculators and spreadsheets would be unable to process all the data being churned out by some of today’s systems.
Kris Hammond, chief scientist of Narrative Science, says that many companies “have spent a tremendous amount on collecting data, and they might have analysts who only produce one or two reports a day”.
Companies with large numbers of branch managers or franchisees, insurance companies with big sales forces and wealth management concerns with thousands of customers are among the businesses his company is targeting for its automated report-writing software.
This points to a technology dividend that often follows widespread automation. Once humans are taken out of the equation and the cost per unit of work collapses, volumes explode. Telephone switches, for instance, enabled a volume of calls that the switchboard operators they replaced would never have been able to handle.
Whether this is good or bad for individual workers will depend on where they stand on the all-important hierarchy of “knowledge” work.
“Technology substitutes for lower- level people and frees up higher-level people to engage in ever more cognitive, ever more cerebral activities,” says Mr Nadler.
For workers worried about their jobs, the critical question will be where that line is drawn – and whether there will be enough of the brainier, higher-level jobs to go around. A second issue also looms large: whether new types of work are invented at a fast enough pace to replace the jobs that are lost.
The history of other technology transitions gives cause for optimism, says Tom Malone, a management professor at MIT and author of The Future of Work. “In every single case where people have worried about that, in the long run just as many jobs were created as destroyed.”
In the century or so leading up to 1910, for instance, the automation of agriculture reduced the proportion of US workers engaged in the sector from 90 per cent to 2 per cent, says Mr Kaplan. While individual workers may have suffered, new productive uses were found for the workforce as a whole.
Yet that does not guarantee the transition will be smooth, says Mr Malone.
The pace of digital change in many industries has been quickening. Twenty years ago, Fidelity Investments, one of the largest US mutual funds groups, conducted almost all of its business over the phone, says Sean Belka, head of its advance technologies division. In particularly busy periods, even headquarters staff were co-opted to answer the phones.
These days, 12,000 of Fidelity’s staff – or three workers out of every 10 – are employed in IT roles and the company spends nearly $1bn a year on new software projects such as smartphone apps. Fidelity is now among the companies testing technologies including Warren and Quill as it tries to anticipate how work will change next.
“Business will just shift to more technologically enabled companies,” predicts Mr John at Rocket Fuel. His company’s staff more than doubled last year to about 600 – a reflection of the wider upheaval that is transforming advertising.
Like airline pilots, more and more people will come to find themselves, at work, as “humans embedded in complex systems”, Mr John says.
If they still have a job to go to, that is.
Robots: Blue-collar workers beware: Baxter is after your job
The robots are ready to move on past their jobs on the production line.
A $25,000 machine with a tablet computer for a face, Baxter – the product of US company Rethink Robotics – is one of the first all-purpose robots designed to move around and handle a range of tasks.
Recent breakthroughs in computer vision, long one of the toughest challenges in artificial intelligence, lie behind the emergence of machines such as this, says Rodney Brooks, the former AI professor who dreamt up Baxter. He also credits a collapse in the price of sensors (another byproduct of the smartphone revolution) and machine learning that supplies the “brains” of the machine.
For now, Baxter is most likely to be found moving bulky packages in a warehouse or loading a truck. Warehouse workers are “nothing but hands and eyes: go to this bin, pick it up, put it somewhere else”, says Jerry Kaplan, a Silicon Valley entrepreneur, who predicts most will quickly be replaced by robots.
Amazon and Google last year bought two of the leading warehouse robotics companies. The involvement of these companies could trigger a technology race that brings rapid advances, says Mr Brooks. It could, for instance, lead to robot “hands” with fine motor skills – another challenging area.
The arrival of robots such as Baxter threatens to upend a long-held belief about work: that many types of low-skilled manual work are simply too hard to automate and will remain the preserve of humans.
Full automation may not be practical, says Mr Kaplan, but when most aspects of a job can be given to a robot, the work is likely to be restructured to let the machines do what they do best.
It may be decades before Google’s pioneering driverless car leads to a fully automated taxi service. But autonomous vehicles have already proved themselves safe in highway driving. That could lead to drone trucks carrying goods on transcontinental highways, handing off to human drivers to quickly negotiate the last few miles, Mr Kaplan predicts.
Many jobs could be rearranged this way, cutting a swath through the blue-collar workforce. “Digging ditches, laying pipes, directing traffic – it’s just going to start knocking them off.”
Letters in response to this analysis:
Get alerts on Alphabet Inc when a new story is published