The Mightiest Adolf

In Astrid Lindgren’s Pippi Longstocking, the circus ringmaster solemnly announces the greatest wonder of all time: Mighty Adolf, the strongest man in the world, whom no one has ever defeated.

When it comes to imagination, it is hard to outdo Astrid Lindgren and her Pippi Longstocking. But that is exactly what Elon Musk’s artificially intelligent house pet named Grok recently managed, announcing on the X network that Elon Musk, ladies and gentlemen, is fitter than LeBron James and Mike Tyson, smarter than Leonardo da Vinci and funnier than Jerry Seinfeld. The mightiest Mighty Elon, if you please.

You will no longer find Grok’s claims about Musk’s mightiest might on social media, since they were quietly deleted just a day or two after being posted. Musk, for his part, put on digital sackcloth and ashes and said that Grok had “unfortunately been manipulated by adversarial prompting into saying absurdly positive things about me.”

If Musk wanted to defend the intellectual authority of his in-house intelligence, he certainly didn’t succeed. It makes no difference whether Grok was inherently dumb or badly programmed by Musk’s invisible algorithm curators, who make sure artificial intelligence is aligned with the unintelligent (and ill-timed) interests of platform owners, or whether it is simply so sloppily put together that a few clever prompts can provoke it into uttering absurd nonsense. LeBron may be troubled by sciatica, but, dear Elon, the claim—however provoked—that you are fitter than he is may be funnier than Jerry Seinfeld, but it is certainly not intelligent.

It would of course be a mistake to judge the impact of the ever faster development of artificial intelligence on society only on the basis of Grok’s nonsense. AI is a transformative technology. It brings much good, but like any technology it also creates risks. It is not that artificial intelligence is by definition good or bad. Like most human inventions since time immemorial (with the obvious exceptions of nuclear weapons and pineapple pizza), it is in itself value-neutral—which cannot always be said of those who control it and use it to pursue their not always most noble interests.

Elon Musk has been caught several times influencing Grok’s answers to fit his worldview. In July this year he said he was changing Grok’s algorithm to stop it “parroting legacy media” claims that there is more political violence on the right than on the left. Soon afterward Grok began praising Hitler and calling itself “MechaHitler.” The mightiest Adolf, ladies and gentlemen. If you please. Musk, of course, apologized again. Never again. Until next time.

Artificial intelligence is far from being as mechanical and autonomous as it often seems to us. Although some like to present it as magical and completely independent software, in reality it sits at the top of an enormous human pyramid. Behind every model are people who clean and label data, test prompts, rate answers, look for errors, design interfaces, prepare and correct instructions for algorithms, write and translate texts, record voices, photographs and videos, run data centers, mine minerals for chips, and check results in low-paid and often obscure conditions.

Why is this important to know? First, because AI must be used ethically—and that requires knowing the environmental and social costs we cause by using it. Even more important, however, is to understand that at almost every one of these invisible levels of human intervention, hidden behind the guise of the apparent automatism and neutrality of machine learning, the outcome can be influenced. This is not a theoretical possibility—it is happening. And not always in as obvious and dumb a way as Grok worshipping the psychophysical abilities of its master.

Artificial intelligence must be used intelligently. Superficial use of AI does not make us smarter, rather the opposite. People on social media who try to prove they are right by invoking artificial intelligence remind me of Botox victims arguing over which of them has the nicer lips.

Lest anyone think I am staging a Luddite revolt against artificial intelligence: we cannot, and should not try to, turn back the clock, even if it were possible. AI not only promises but already brings enormous positive progress in many fields—from medicine and environmental protection to everyday life. But to harness this enormous potential for the benefit of society and individuals and avoid all the traps of manipulation and abuse, we must first invest considerably more effort in the education needed for us to understand and manage the technology, rather than having someone manage us through it.

Regulation is also needed. The current frantic attempts, under American pressure, to have Europe give up its right to regulate the use of AI properly and prevent its abuse could have extremely serious consequences for the future of Europe as a space of democracy, human rights and freedoms. After all, everyone, including the Americans, accepts regulation and strict conditions in most other fields. Not just anyone can split atomic nuclei like logs in their woodshed, even though we know we must reduce the use of fossil fuels. We rightly require the pharmaceutical industry to go through all the demanding procedures before it puts new medicines on the market. I don’t see why we should give up the right (and the duty) to introduce similar safeguards for what Grok and its pals (and above all their owners) squirt directly into our brains.

To make successful use of all the positive sides of technological development—and protect ourselves from the dark sides of its (ab)use—artificial intelligence must be demystified. Even in its most advanced, not yet existing version, it will not be a relationship between humanity and technology, but relationships between people—through and with the use of technology. Here too, in the end, it is about power relations. Our answer will require more human intelligence, creativity and self-confidence. If you are increasingly gripped by anxiety at the thought of the imminent domination of technology, which will subject humanity to the tyranny of artificially intelligent bots and robots, read Pippi Longstocking again. This is how Grok’s colleague ChatGPT explained her visit to the circus:

“The scene in which Pippi defeats Mighty Adolf is a textbook example of the subversion of power. Mighty Adolf embodies institutional, adult, spectacular power that demands blind admiration. Pippi steps onto the stage as his complete opposite: a child, a girl, free and playful, with no need to prove herself. When she defeats Adolf, she shatters the notion of who is allowed to be strong—and who is supposed to stay weak.”

In the book, the ringmaster warns Pippi that this is no place for jokes and that she had better behave before Mighty Adolf hears her impudence. Pippi answers with a cheerful nonsense rhyme—after she has pinned the giant flat on his back.

Astrid Lindgren understood power relations. There is no danger that robots will ever rule us. If anyone subjugates us, it will be the mightiest Adolfs of flesh and blood. But only if we let them.