Underpriced AI models could trigger mass unemployment, pioneer warns

AI researcher Bernhard Schölkopf fears that rapidly advancing and underpriced AI models could eventually eliminate large numbers of jobs, Yle reports.

“It is clear that we are now in turbulent waters,” Schölkopf says. The German researcher is one of the world’s leading specialists in artificial intelligence and machine learning. He is a professor at ETH Zurich in Switzerland and director of the Max Planck Institute for Intelligent Systems in Tübingen, Germany.

Schölkopf has witnessed the development of machine learning, the technology underpinning today’s AI models, into functioning systems that are expected to transform the world. He says the pace of change has made him more humble.

“Development will probably proceed in a very different way from what even the smartest people are able to expect,” he says.

So far, AI has had only a limited effect on the labour market. Schölkopf believes that continued technological progress could change this quickly.

“Sometimes gradual progress reaches a point where we suddenly realise that the system is so good that something completely different can suddenly be done,” he says.

He says programming has reached such a point. AI systems can now produce usable code, making it economically viable to undertake projects that would previously not have made financial sense.

Schölkopf compares this development with the impact of information technology, which enabled industrial-scale knowledge work from the 1970s onwards. AI, he says, makes it possible to automate that work by processing large bodies of information that could not previously be used in practice.

One example is Google’s 10 million dollar purchase of the database of bankrupt US airline Spirit Airlines. Google bought the airline’s emails, internal messages, flight booking data and payment receipts to train its AI systems.

“Ten years ago, data of this kind had practically no value. Now it can be used to build an AI system that understands the airline’s internal operating mechanisms,” Schölkopf says.

“I feel that in programming we have only reached this point this year. If that is true, the effect on the labour market will only appear with a delay. Even though economists do not yet see any effect, I am concerned that the impact will be quite significant.”

Europe must influence AI development

Schölkopf says the ongoing development cannot be stopped. But he does not believe this means society must simply accept whatever follows.

“It does not help to put on life jackets and just hope that we will not drown,” he says.

In his view, the best way to stay afloat is to influence the development of the technology itself. He says Europe has suitable mechanisms for doing so through its open society and democracy.

EU Commissioner Henna Virkkunen, responsible for digital affairs, also believes in this approach. Her task is to strengthen Europe’s technological self-sufficiency.

In recent years, AI research has shifted increasingly from universities and public research institutions to the research laboratories of private companies. These companies are mainly American and increasingly Chinese.

“This means that AI research as a whole is moving away from Europe. If AI really reshapes our world and society, it is worrying that the change is being led from elsewhere while we are merely passive consumers and bystanders,” Schölkopf says.

Schölkopf has helped establish the Ellis Institute, a research centre and network for European AI research. The network expanded to Finland last year, when researchers employed by the institute began working at universities around the country.

The institute also brought Schölkopf to Finland in June to give several lectures in Otaniemi. He believes an ecosystem could emerge around the Ellis Institute in Finland, allowing growth companies to flourish.