The spread of AI systems in the world of work has been debated with either enthusiasm or unease for a number of years. The Federal Ministry of Labor (BMAS) took the debate so seriously that in March 2020 an “AI observatory” was launched in the ministry’s “think tank”. Together with the University of Düsseldorf and the Center for Advanced Internet Studies, the observatory interviewed around 600 employees about the changes they fear most from AI systems. The most important answer is surprising: “The majority of respondents do not expect to lose their jobs in the next 20 years,” says BMAS State Secretary Björn Böhning. In comparison, surveillance in the workplace or a lack of transparency about what happens to personal data is more of a concern for people. On the other hand, many of them are pinning their hopes on increasing occupational safety and relief.
This article is from issue 04/2021 MIT Technology Review (order as pdf). As a special issue, the issue deals with the future of work.
As recently as 2013, the Oxford researchers Carl Benedikt Frey and Martin Osborne warned that 47 percent of US jobs would be lost as a result of AI applications. And the Israeli historian Yuval Harari wrote three years later in his book “Homo Deus”: “The most important question of the economy of the 21st century may be what we do with all the superfluous people.”
There is much to suggest that things will turn out very differently. Because after the initial hype about the amazing capabilities of deep neural networks and machine learning algorithms, it is becoming increasingly clear how limited their capabilities are. Because the underspecification of training data or unintended shortcut conclusions in neural networks sometimes make these algorithms unreliable. Above all, however, they do not produce any really new ideas. On the other hand, if you combine AI systems with the experience and creativity of real people, productivity increases. “It’s not about computer models against people, it’s about computer models together with people,” says Menno van der Winden, general manager for advanced analytics at steel manufacturer Tata Steel Europe, who himself relies on AI systems.
The vision of the hour is called hybrid intelligence. Instead of freeing the flow of data between sensors, databases and machines from human hands, eyes and ears, people are deliberately integrated into the work process – in English: “human in the loop”. “Hybrid intelligence is the combination of human and artificial intelligence to augment human intelligence instead of replacing it,” says the Hybrid Intelligence Center, which was founded in 2019. Leading AI researchers from the Universities of Delft, Amsterdam and Groningen are working there on the future of AI.
Examples of such a combination are becoming increasingly common in business. But what exactly do they do? How do humans and AI complement each other in practice?
To home page