It doesn’t happen every day that you suddenly have a son. Or a brother. A husband. Or a daughter who follows in your footsteps and also writes columns.
Just to be clear: I have no brother, no husband and no son. And my daughter doesn’t write columns. Yet I read that recently, when I Googled myself in AI mode. “Japke-d. Bouma has a brother, Rudy Bouma, who is a correspondent for the NOS. She lives with her husband in Utrecht. Her daughter follows in her footsteps and also writes columns.” A friend texted that Google had told her that I have a son, and that my daughter’s name is Doutzen Bouma, “named after her mother.”
That wasn’t right either.
It’s all made up. Through Google’s ‘AI mode’. You know. The artificial intelligence mode that no one asked for. Who quickly sifts through millions of websites and texts. And it costs tons of electricity and liters of water. Super sin. Because what comes out is often wrong. They call that ‘hallucinating’.
I was pretty pissed off that he had made me into a lame hat. And not that I have a secret relationship with Brad Pitt, am a winner of the NS People’s Choice Award or have ever been in a working group for world peace with Barack Obama.
But it also scared me. Because what if he had made up for me that I had stolen the cash at NRC, sold my child on a slave market or eaten my cats? Everyone would read that too.
And so Google, eh!? The search engine that used to – usually – neatly lead you to Wikipedia or a news site. But now in AI mode you suddenly start creatively inventing human lives. Would everyone know that?
‘According to ChatGPT I was already dead’
Artificial intelligence still has some, er, start-up problems anyway. Take ChatGPT. A friend recently asked ‘him’ to send a PowerPoint to his email address. It took hours. Chat kept running. ‘I need a different email address’. “It’s really happening now.” “I know I don’t deliver what you asked for, but I’m going to make it right now.” “This really has to be better, this is not what you expect from me.”
Until the monkey came out of his sleeve, and turned out, because that was technically too complicated for him. He had deliberately created false email addresses, excuses and excuses that made no sense. And when that came out: even more apologies.
That is recognizable, as it turned out on LinkedIn. People who, according to Google AI, suddenly live in Toronto. People whose son is their husband. Bus lines that do not stop at the desired location. Seniors who are suddenly government trainees.
There was also someone who had Chat draw up minutes that used completely randomly selected minutes from a meeting elsewhere in the world. Someone who, according to Google, was a fictional character “created by Herman Koch” – I would have liked to be that too! To top it all off, the person who wrote: “I have apparently already died. After a short illness. Fortunately, my associates continued my good work after my tragic death.”
ChatGPT? That’s ‘KletsGPT’
Jokes were made about it on Linkedin. Professor of Organizational Change Woody van Olffen calls ChatGPT ‘chat GPT’ – we’ll keep it that way. “Maybe Chat sees more in you than yourself,” wrote someone. “It is also human that Chat does not dare to say: ‘I don’t know’,” another.
“Chat looks a bit like one of those dates who acts like he has everything under control, but when you get to his house the whole kitchen counter is full of moldy dishes,” wrote a friend. Or an uncle who ‘read something on Feestboek’ on a birthday. “The delaying, the lying, the twisting, the promises, the excuses and the excuses,” my friend wrote. “I’ve honestly never seen this before.”
I do, I thought. AI is starting to look more and more like a colleague. But one of those that you would rather not have too many of. Such an irritating slimeball, who lies flat out, and then starts to turn around when he really has to deliver.
AI search engines should say: I don’t know this. But they are designed to please. This produces fake news, which is taken over and pumped further by other machines. That makes me fearful for the future in which more and more people will rely on that kind of information.
AI is a deadly nuisance force
A number of readers pointed out to me on LinkedIn that AI models are improving quickly. That you can set in your ‘prompt’ that no information may be made up (really?? It doesn’t do that automatically?!). That it would be a shame “to let fear get in the way of much-needed innovation.” But I thought: this is not a much-needed innovation, this is a life-threatening nuisance. If this is the future, we need to become a lot more careful.
Because artificial intelligence is not a colleague who can be fired if he messes up again and again. AI is a black box that we will never get rid of. A highwayman who comes back with stuff that we have no idea where it came from.
Because making up a brother – if only it were so innocent. Artificial intelligence steals information from behind paywalls. Talk to the Russians. Undresses women, children and babies on Twitter. Advises people to commit suicide. Puts texts in the mouths of marriage officials, making marriages invalid. And those are just a few qualifications I came across on the web. But don’t worry, man.
Maybe that’s not right either.
The journalistic principles of NRC