AI & Innovation: The Good, the Useless – and the Ugly

“Boosting innovation” is the new rallying cry in EU AI policy. But the frenzy overlooks that while some forms of innovation are beneficial, others range from unnecessary to harmful. Pursuing AI innovation indiscriminately threatens to waste resources and energy on developments that serve no purpose or, worse, negatively affect society. As Europe considers its options to achieve digital sovereignty, it is time for a sober assessment of the prevailing innovation narrative.

For over two decades, the Commission wrote in its Competitiveness Compass report, the EU has fallen behind other major economies. Low productivity has caused anemic economic growth, and the “root cause” of it all is “a lack of innovation.”And if there is one field in which the EU must innovate, AI is it.

Innovation is the kind of “what’s not to like?” buzzword, like trustworthy AI or AI ethics, that attracts applause from all EU quarters. And what could be said against more innovation?

A few things, it turns out. In particular, the hype around AI rests on the idea that innovation will drive productivity, which will drive growth and strengthen competitiveness, ultimately improving welfare. The argument sounds sensible enough. But there are enough shortcuts in each link of the argumentative chain to justify putting it under a magnifying glass.

AI and innovation: the good, the useless, and the ugly

First things first: there is good innovation, the kind that genuinely benefits society. AI that enhances energy efficiency in manufacturing, aids scientific discoveries, improves extreme weather prediction, and optimizes resource use in companies falls into this category. Governments can foster those innovations through targeted R&D support, incentives for firms to develop and deploy AI, “buy European tech” procurement policies, and investments in robust digital infrastructure. The Competitiveness Compass outlines similar strategies.

That said, given how many different technologies are lumped together in the AI category—everything from facial recognition technology to smart ad tech, ChatGPT, and advanced robotics—it makes little sense to talk about good innovation and “AI and productivity” in the abstract.

Most hype these days is about generative AI systems that mimic human creative abilities with striking aptitude. Yet, how transformative will an improved ChatGPT be for businesses? It might streamline some organizational processes, expedite data processing, and automate routine content generation. For some industries, like insurance companies, such capabilities may be revolutionary. For many others, its innovation footprint will be much more modest. Meanwhile, other forms of AI–particularly in robotics, where robots are often trained “on the job” with reinforcement learning, may transform manufacturing and other sectors that far exceed the capability of generative AI.

The distinction matters. Many of the most productivity-enhancing forms of AI don’t require data centers that cost hundreds of billions of dollars to build and manage. Much smaller, more efficient systems will do. From a productivity perspective, the race between DeepSeek and OpenAI is a big distraction. The huge gap in “AI investment” between the US, China, and Europe contains a lot of money spent on technologies likely to have limited economy-wide impact.

Still, generative AI remains a valuable endeavor for today’s tech giants. Just days after the DeepSeek announcement, Zuckerberg boasted that he’d double down on AI with “hundreds of billions of dollars” to invest in Meta’s AI. The company’s stocks rose in response. Why? Because investors rightly sensed that Meta would be able to target ads to people even more effectively.

While some advertising delivers useful information, much of it is about creating material desires—distractions from building deep social relations that truly make people happy once basic material needs are met. Elsewhere, AI is used to extract value from small businesses and workers, such as algorithms that squeeze sellers on the Amazon marketplace. Moroever, if AI removes human interaction from our economic exchanges, it makes them more sterile and less meaningful, both for workers and for customers. Jobs become less enjoyable and more stressful; ultimately, society bears the cost.

In short, a European innovation system that concentrates public support on genuinely beneficial use cases—not just things that make money—is a much more effective and efficient way of using resources. That doesn’t mean governments are picking winners. It only means supporting companies and research areas that promise to deliver real value to society. AI outside that category can be safely ignored, no matter how lucrative it might be to the tech industry.

GDP growth + competitiveness = societal progress?

Lackluster GDP growth in Europe is a key plank of the innovation hype. The Draghi Report presents it as central evidence that Europe is falling behind. The assumption is that robust growth undergirds societal progress. (If it didn’t, why should we care?)

But many things that count as growth don’t make societies better off, and vice versa. A few examples: In the bloated US healthcare system, vast sums go toward managing public health crises like diabetes. In 2023, more than $80 billion were spent on diabetes drugs, roughly half of which went to variants of Ozempic, which is also popular as a weight loss treatment. Is a healthier society that doesn’t need to incur such costs really poorer?

Meta, mentioned earlier and the world’s largest advertising company, typically makes between 45% and 50% of its overall revenue in Canada and the United States—in 2023, it generated a little more than $61 billion in those two countries. The company’s booming ad business and the extra sales it drives boost North American GDP. But if AI tempts more people to buy more clothes, more frequently, only to throw them away sooner, is that progress– or is it just wasteful?

Or consider a country in which people have to work so many hours that they need to hire others for daily tasks: People to mind the kids, people to care for elderly parents, people to deliver food and groceries, and so on. These transactions inflate GDP on paper, even though, in practice, it leaves many people exhausted and unhappy. Again, the United States stands out here, with annual working hours higher than all but two European OECD countries. ( On average, Germans work 25% fewer hours than those in the US—time spent on valuable leisure or family, for example.)

Indeed, focusing on GDP growth alone can undermine other policy goals and societal benefits. Imagine a place that develops great public infrastructure (or bike paths everywhere, like here in Amsterdam), meaning fewer people need to buy and own cars. That scenario technically shrinks the economy—even though productivity has actually increased: greater mobility with fewer vehicles and less environmental damage.

Long story short: there is both good and bad GDP growth. But in the Excel sheet—or in the graphs that dot the Draghi Report—you can’t tell them apart. We shouldn’t worry about lower headline figures in the EU as long as we do get the growth that is worth wanting.

Many champions of competitiveness still imagine a world in which cost competitiveness decides triumph or defeat in a global marketplace. Tariff barriers have already risen in recent years, and Trump’s punitive levies on Mexico and Canada may have delivered the final nail in free trade’s coffin. Making more things with less is always great, whether you trade with others or not. But the argument that your survival depends on being cheaper than all the others no longer holds. If, say, your cars are a bit more pricey because you pay workers fairly, that may be just fine.

Both the Draghi Report and the Competitiveness Compass make the case for reducing excessive dependencies and increasing security. That strategy makes perfect sense. Trump is happy to bully his closest allies, and Europe has no reason to expect different treatment. If nothing else, EU dependence on US tech is an invitation for extortion from the White House. Alternatives, as proponents of a #Eurostack have argued, are of the essence.

That, too, requires targeted support for critical systems. If Candy Crush (owned by Microsoft these days) disappeared from iPhones tomorrow, no real harm would be done. If X vanishes, the last holdouts could switch to BlueSky and shrug. If, in contrast, Microsoft withheld the next batch of MS Teams security patches, thousands of companies and millions of citizens would be dangerously exposed.

Europe doesn’t need to blindly embrace every aspect of AI or shape its innovation policy around the idea that survival depends on leading the global race–or, as European Commission President Ursula von Leyen emphasizes, “to succeed in the race to the top.” The lodestar for European AI policy should not be misleading GDP figures or outdated ideas about competitiveness but maximizing the social utility of technology. That’s an innovation goal worth pursuing.