Unless you’ve been hiding out in a doomsday bunker you’ll have heard the news that the robots are coming for your job.
They are already delivering takeaways in Washington, DC, flipping burgers in California, doing data entry in accounting firms worldwide, writing articles for the Associated Press, appealing parking tickets, correcting errors on Wikipedia, managing Amazon’s warehouses, creating ads for Coca-Cola and helping in complex surgery.
Soon “there will be fewer and fewer jobs that a robot cannot do better”, according to the gloomy prediction of Elon Musk, the Tesla and SpaceX chief executive and notorious worrier about the robot apocalypse.
Lest you think your role is safe, a recent report by the McKinsey Global Institute brings the cheery news that by 2055 robots will be doing more than half of what qualifies as work in today’s world.
McKinsey has provided a search tool where you can check how vulnerable your occupation is to partial or complete automation, but I’ll save you the trouble: unless your job involves owning the corporation that builds the robots it’s probably deemed at risk.
You can’t open a newspaper or a browser any more without reading about how we’re on the cusp of an automation apocalypse. The Bloomberg Intelligence chief economist, Michael McDonough, recently tweeted a graph showing a huge surge in the use of the words “artificial intelligence” in the transcripts of earnings calls held quarterly by publicly listed companies.
If you strip away the hyperbole, though, you’re left with a couple of alternative visions of the state of artificial intelligence in 2017. The first is one that looks less like an army of robots coming for your job and more like a coterie of Silicon Valley executives desperately trying to figure out ways to capitalise on the hype.
The second is of the gradual sharpening of a few worthy but ultimately fairly dull technologies, particularly automatic translation, image recognition and decoding the patterns of language.
Take the much-publicised news that Facebook intends to use artificial intelligence to detect suicidal tendencies in posts, in an effort to combat the hugely distressing phenomenon of suicides being live-streamed on the social network. Recently a 12-year-old American girl filmed her suicide live on another site, and it was widely shared on Facebook; in another a 14-year-old took her own life after a two-hour Facebook Live session.
According to news reports, “posts will be compared to others that warranted intervention and, in some cases, passed on to Facebook’s community team for review. People watching a Facebook Live broadcast will be able to report videos for an escalated response from the company.”
This sounds not so much like a leap in artificial intelligence as simply a variation on the system that has previously been used to remove, for example, photographs of breastfeeding mothers.
Facebook chatbots
Last year enormous hype greeted Facebook’s announcement that it was launching chatbots in Messenger. Somewhat less well reported was the admission a few months later, by Facebook Messenger’s vice-president, David Marcus, that the technology “got really overhyped very, very quickly”.
Marcus said that the “basic qualities we provided at that time weren’t good enough to replace traditional apps”. One of the apps launched on stage at the Facebook announcement was the weather bot Poncho.
I interacted with Poncho last week, almost a year on, and although it was unable to geolocate me on a map it did provide me with the weather for my area once I typed in my location. (It also advised me to wear gloves and snow boots in cloudy, 10-degree conditions.)
Another example of the benefits of artificial intelligence is Google's machine-learning tool for publishers. Called Perspective, and developed in conjunction with the human moderators of the New York Times, it will help online publishers identify and remove "toxic comments".
Once again, I tested the tool online with mixed results. The comment “You are a man” was deemed 22 per cent likely to be toxic, while “You are a woman” had a toxicity reading of 30 per cent. “You are a feminist” was 65 percent likely to be toxic.
Another manifestation of artificial intelligence that, I think, many of us could get behind is the creation of a chatbot, or “robot lawyer”, by the Stanford University student Joshua Browder. The Do Not Pay service, which offers “free legal help in under 30 seconds”, appeals parking tickets in New York and the UK and delayed flights or trains in the European Union.
In theory this kind of process lends well itself to artificial intelligence: it is time consuming but highly structured, with a finite set of possible responses. Browder says his service has been successful an extraordinary 64 per cent of the time, overturning 160,000 parking fines.
Again, I tried it out on a fictional delayed train from Waterford to Dublin, and, again, there were a few hitches: the chatbot autogenerated a letter of complaint that referred to my train as a flight, for example, and claimed I had an address in the UK.
The student has since turned his attention to helping to provide housing for homeless people in London. “The value in bots is not to order pizzas,” he said recently. “The possibilities are endless.”
At least for now, though, artificial intelligence has limited applications. And although, ultimately, it is inevitable that some jobs will be made obsolete by automation, new jobs are almost certain to be created by it, too.
What kinds of jobs they might be is difficult to say. After all, one in three new jobs created in the past 25 years in the United States didn’t exist at the beginning of that period.
In a report on artificial intelligence and the economy during the Obama administration the White House suggested that jobs could be created in supervising artificial intelligence, repairing systems and reshaping infrastructure to accommodate things such as self-driving cars.
The robots are coming, it’s true. But I wouldn’t start campaigning for a universal basic wage or building that bunker just yet.
Know your chatbots: a tech glossary
Artificial intelligence: computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decisionmaking, and translation between languages. Technology that learns, then acts.
Machine learning: the science of getting computers to act without being explicitly programmed. Applications include self-driving cars, speech recognition, effective web search, and a vastly improved understanding of the human genome.
Deep learning: a branch of machine learning focused on complex and large-scale data sets.
Bot: a web bot, or internet bot, is an automated program that runs simple and structurally repetitive tasks at high volume over the internet.
Chatbot: a computer program designed to simulate conversation with human users.
Robot: a machine capable of carrying out a complex series of actions automatically, especially one programmable by a computer.
