Learning by doing – Catching up with AI and technology developments

by Fredrik Erixon

Metaphors sometimes lead us astray. Right now, we are told, an AI revolution is unfolding, one that will “change everything".?But what if it isn’t a revolution but an evolution - an escalating, yes, but gradual and sometimes slow-burning development following the logic of Charles Darwin rather than the hot-temperament revolutionistas?

Very few things actually change everything, and most technologies that come to power our future economic development evolve over time in a process of learning and adaptation, trial and error. So what Europe should worry about is not that we’re not missing a revolution, but that too few in our part of the world have made our entry to that evolution.

I think many of Europe’s policymakers are flippant about the risk of technological backwardness. In a newspaper interview this weekend, Brando Benifei, who shepherded?the AI Act through the European Parliament, dismissed Apple, Meta and others who have not released new AI tools in Europe. They will launch in Europe, he said, “it is only about a few months, perhaps a year”. Margarete Vestager, the EU’s outgoing competition czar, doesn’t seem to mind either that the new AI functionalities on Apple’s new iPhone won’t be part of the European release for a while. She is rather “quite relieved” that the new AI functionality isn’t available on her iPhone.

So what’s the problem? Well, innovation delayed is innovation denied. You may not want AI in your device, but others may do. Innovation is an evolution, and when we in Europe eventually get access to technologies, these technologies will have evolved, and perhaps something new and better will be introduced in other places. It takes time for users to build up knowledge and habits, in the first place, just to understand what new technology can do for you. Fears of new technology – and, certainly now, with AI – often come down to inexperience and not knowing what it’s all about. Innovation, of course, is much about what you stop doing in order to make place for something new. If you don’t have access to frontier tech, you won’t even know where to start.

There’s another economic angle to this: human capital needs an upgrade with new technology, and delayed knowledge development for labour make more of us at risk of not really being part of a new development. Some of my colleagues pointed this out in a study we released last week about economic opportunity and improving credentials. Working with new technologies is about skills, not titles, and there is a giant economic boost waiting for Europe if we could get a lot more workers basic training in technology. It’s not about taking factory workers into a four-year university degree in data science: it’s about short, focused, and structured vocational learning.

The labour market is often missing when Brussels ponders new ways to deal with risks of technological backwardness. We released last month a study taking stock of technology diffusion in Europe (a fancy word for tech take-up), and it makes a similar point: yes, there are several structural hinders to tech spreading faster in the EU, but fixing it is to a very large degree about skills – and improving them. There has been strong convergence within the EU already in basic indicators of technology readiness: what’s needed is not new regulations but a stronger focus on skills.

Reading these papers by my colleagues, I was reminded of my grandfather. He was old when the ICT revolution started to unfold in the early 1990s, and he didn’t have much experience with computers. But he took an evening course in basic computer skills and, using Excel, soon started to digitalise his own records of weather observations, going back to the 1940s. He systematised the data and found discrepancies in official records, and could help a researcher in historical meteorology with a failing database. I turned to my grandfather for help when I took my first class in econometrics in 1995. He was close to 80 at that point but taught me the basic skills of working efficiently with data in a structured data set.

No, this isn’t rocket science. It’s about learning by doing.