AI & jobs: One door closes and another opens
In 1978, the German magazine Der Spiegel predicted a “social catastrophe” brought on by the computer age. The dire warning included a trade union leader’s declaration that by the year 2000 most jobs in the UK would have been replaced by computers.
As it turned out, the unemployment rate in the UK in 2000 remained at 6 percent, just as it stood in 1978. Today it is even lower, at 3.8 percent.
“Every single time in the past, when technologies have taken away human jobs, more jobs have eventually been created than were eliminated,” explains Thomas Malone director of the MIT Center for Collective Intelligence, in Samsung NEXT’s video series End of the Beginning.
“Take the printing press,” Malone says. “That eliminated jobs for scribes who used to spend their days copying books by hand. But by making it dramatically cheaper and easier to copy text, the printing press basically created the entire modern publishing industry.”
A more contemporaneous example is Google. The rise of the search engine may have led to a decline in jobs for reference librarians, but it has also enormously expanded employment opportunities for everyone from programmers and database manager to website designers and search engine optimization specialists.
From logical reasoning to emotional intelligence
Still, the prospect of retraining for these entirely new roles may seem small comfort to those working in fields deemed ripe for human-computer substitution. And it’s no longer just telemarketers, accountants, and loan officers who have reason for concern.
As machine learning techniques grow increasingly sophisticated, even computer programmers are beginning look vulnerable to replacement by their own creations.
In the traditional approach to artificial intelligence (AI) — responsible for systems like IBM’s chess-playing wunderkind Deep Blue — humans derived and encode the rules of intelligent behavior by hand.
By contrast, today’s machine learning algorithms automate the process of uncovering the relevant regularities of a domain — whether that’s the necessary features that make up a photo of a cat, or the optimal strategy for the video game StarCraft. To do so, they require nothing more from human trainers than feedback on whether the current result is correct, allowing the system to repeatedly adjust until it hits on a reliable strategy.
“Thirty or 40 years ago, the people who actually programmed computers were computer scientists with prestigious degrees, working in a clean room full of big IBM machines,” says Yosi Taguri, whose company MissingLink.ai designs software to bring machine learning to the masses. “Today you don’t need any of that.”
The ability to learn regularities, rather than relying on hand-coded rules, is also helping AI expand beyond its traditional domain of logical “if-this-then-that” reasoning into the messier human realms of visual, emotional, social, and even creative intelligence.
“I think we’re going to see machines moving beyond tasks like just vacuuming the floor,” says Dor Skuler, CEO and co-founder of social robotics startup Intuition Robotics. His company’s first product is a home robot called ElliQ that is designed to act as a companion for socially-isolated elderly people by learning their behavioral patterns and responding in personalized ways.
“We’re going to see the voice assistant evolve into something much smarter that uses context and emotional intelligence to adapt to our personality,” Skuler says. “Into something we can actually build a relationship with.”
The problem of AI
From learning to outperform human masters of the notoriously abstract strategy game Go from scratch in a matter of just three days, or using facial recognition to evolve and adjust adverts in real-time in response to viewers’ reactions, the capabilities of cutting-edge machine learning certainly look impressive. Today, developers may have reached a tipping point where even the new work created by AI advancements can be performed by AI systems themselves.
Still, MIT’s Thomas Malone isn’t too concerned about the evolution of AI systems. Despite the strides of the past few years, there’s one crucial skill in which machines aren’t even beginning to approach human capacity — general intelligence.
“If we see a computer program playing chess at a level that can defeat a world champion chess player, we assume that that program must be as smart as the human chess player, but in reality the program can’t begin to do most of the other things that a human chess player can do,” Malone says. “Even a 5-year old kid can carry on a sensible conversation about a much wider range of topics than the most advanced computers today, and can operate much more effectively in an unpredictable physical environment.”
From competition to collaboration
Rather than continue to worry about how, when, and where AI will replace humans, and what can be done about it, Malone advocates looking at things from a different angle. To shift the conversation from how humans compete with AI at a particular task towards considering the entirely new capabilities that can be opened up when they work in collaboration.
Computers have already enabled collective intelligence at a previously unimaginable scale, he points out. Take the example of Wikipedia — an encyclopedia in which nearly 6 million articles are created and constantly updated by an internationally distributed group of hundreds of thousands of volunteer editors. Such an achievement would have been inconceivable for small locally-connected and centrally controlled team.
Now imagine, Malone asks, what will be possible when computers go beyond facilitating large-scale human collectives, to becoming active participants, able to contribute their unique capabilities to an integrated human-computer workforce.
“I think it’s possible that we can create genius organizations,” he suggests. “What you might call ‘super-minds,’ made up from groups of people and humans working together in ways that are smarter than anything that’s ever existed before and will be able to solve our problems far more effectively than we can today.”
Learn more about trends in AI by watching the End of the Beginning video series.