Outgoing Minister Robbert Dijkgraaf (Education, Culture and Science, D66) and State Secretary Alexandra van Huffelen (Kingdom Relations and Digitization, D66) take a seat next to each other on a cramped, gray two-seater bench. They are sitting in the ‘executive lounge’, on the ninth floor of the Marriott Hotel in The Hague.
The hotel is less than a five-minute walk from the Catshuis, where the cabinet was updated on artificial intelligence (AI) in a ‘Catshuis session’ on Wednesday morning. The administrators will speak for an hour afterwards NRC about the technology, which has been on the rise since the launch of ChatGPT in November last year.
The spokesperson present notes during the conversation that it is appreciated if it is explicitly stated that AI was also used for writing this article. “To make clear to your readers how all-encompassing this technology is.”
That’s right. The interview recorded with an iPhone was transcribed by Whisper [software die binnen een minuut een audiogesprek in geschreven tekst kan omzetten]. And on the site of NRC the article is read by a vote of one NRCjournalist trained using artificial intelligence. Here too – as is happening or will happen in most professions – artificial intelligence is taking over tasks from people.
In the Catshuis, the cabinet, at a large oval table with caretaker Prime Minister Mark Rutte in the middle, was updated by, among others, AI professor Max Welling from the University of Amsterdam and cyber expert Marietje Schaake from Stanford University. Discussion topic: the opportunities and risks of AI. It was the first time that the government had put this subject on the political agenda in this way.
Think tank and advisory body WRR mentioned AI earlier a ‘system technology’: a technology with a large, unforeseen impact on society. This concerns AI systems that are used for facial recognition and fraud detection or for medical or military applications. And algorithms used to influence what citizens buy, read and vote. According to the WRR, the technological impact of AI is therefore comparable to that of previous revolutions such as electricity or the combustion engine.
According to Dijkgraaf, originally a naturalist and mathematician, the AI revolution goes even one step further. AI “actually tries to mimic our thinking,” he says. That makes AI something “that touches many more elements of life and is therefore difficult to compare with other technological breakthroughs”.
What did the Catshuis session yield?
Van Huffelen: “There are a lot of sectors in which the Netherlands is strong in which AI will play a major role. Think of agriculture, health care, peace and security, the medical world. The question is: how do you ensure that this technology also incorporates the values that we advocate. The feeling during the session was: we really have to get to work, together.”
Have concrete decisions been made?
Dijkgraaf: “We drive very fast, but through the fog. That makes it difficult to look ahead. The term ‘rapid response team’ was dropped. A group of experts that keeps track of what we are seeing happening here for the government and can respond quickly. We’re going to work on that. ChatGPT was launched less than a year ago. It really goes from month to month now. And that is generally not the pace at which the government works.”
Why has it taken so long for the cabinet to meet on this subject?
Van Huffelen: “Well, of course we have been talking about this theme for some time. But indeed, finding the moment to sweep everyone together…”
Dijkgraaf: “Developments in AI are going faster than the agendas of ministers and state secretaries allow.”
How do you keep up with yourself? Do you use ChatGPT? Do you ever experiment with image generators Dall-E or Midjourney?
“Van Huffelen: “I certainly experimented with it, yes. I read a lot. Listen to podcasts, get inspired by people who are often involved in this subject. But in the end, we can only take in so much.”
Dijkgraaf: “One of mine guilty pleasures is still listening to online science lectures. Then you see, and I find that extremely interesting, all the expectations that we have at Princeton [Dijkgraaf was van 2012 tot 2022 directeur van het Institute for Advanced Study in Princeton] two years ago had now been surpassed. It is precisely this leap that is now being made that blows all researchers off their seats. In science we call this an ’emergent phenomenon’. You are trying to do something. You give it more and more computing power. Nothing happens. Nothing happens. Nothing happens. And at some point everything happens.”
Are your departments already experimenting with AI? Are there working groups or official chat assistants?
Van Huffelen: “That is another question we have. What do we actually want with this? It’s nice to experiment and see what happens, for all I care about answering parliamentary questions. But the question is whether we should want that.”
You said in a speech this week that Big Tech needs to be “tamed” when it comes to AI. How do you envision that?
Van Huffelen: “If these companies want to offer their products to us, they must comply with our rules. Europe is now hard at work on AI regulation [de AI Act, een Europese wet]. It is always the case that technological developments precede regulation. At the same time, we don’t want data from anyone but anywhere to be sold without you having any influence on it. We want regulation in a way that stimulates innovation and at the same time ensures that our public values remain protected here.”
Have you yourself ever fallen victim to AI-generated fake news? A deep fake or a cloned voice?
Van Huffelen: “Not that I know of, but I have seen a lot of misery on social media.”
Dijkgraaf: “I don’t know what I don’t know. So I can’t say with absolute certainty that this never happened or will happen.”
With elections approaching, that should be a major concern
Both, at the same time: “Certainly.”
How do you deal with that?
Van Huffelen: “It is always a difficult balance between freedom of expression and what is really not allowed. We do not want to play the Ministry of Truth or determine what is allowed on social media.”
Dijkgraaf: “There are algorithms that not only create disinformation, but also help you spread it. Chatbots that help you with the intermediate steps. That is worrying.”
Van Huffelen: “This is something that is difficult for us to combat. What we can do is make sure that there are enough resources for people to turn to. Where people can find information that has been tested and where topics are discussed from multiple angles. Media play an extremely important role in this.”
All threats that are already there will be made even stronger by AI
Robert Dijkgraaf caretaker Minister of Education, Culture and Science
Dijkgraaf: “Misinformation is, in a sense, timeless. That will only get worse. That is also one of the most important conclusions of what this technology will do. Any threats that are already there are getting stronger. Then it is also about social inequality. The idea that every new technology is a social equalizer is not correct. All the research shows that this is not the case, provided that you as a government actively intervene. So if you let those forces run wild, the consequences of this technological progress will be very uneven economically and financially.”
I want to present you with some dilemmas. Facial recognition in public space in the Netherlands. May or may not?
Van Huffelen: “In Europe we have indicated that we do not want that. Except under very strict conditions, for example during detection and investigation.”
Civil servants using ChatGPT to answer parliamentary questions.
Van Huffelen: “I am sure that there are civil servants who are already experimenting with this. We need to think carefully about this.”
Dijkgraaf: “Today during the Catshuis session someone said: ChatGPT is just a very smart trainee. You don’t blindly let them sign a letter to Parliament or a contract.”
Military AI systems that autonomously identify targets and then disable them.
Van Huffelen: “No, we certainly wouldn’t want that. I would find that totally irresponsible.”
A six-month break during which all technological development within AI is halted. To understand the technology first.
Both Dijkgraaf and Van Huffelen burst out laughing when they hear the question.
Van Huffelen: “That is simply not realistic.”
Dijkgraaf: “It is now going much too fast for that.”
Yet. There are top scientists who say: this may be the last invention that man will ever make. Tech entrepreneurs like Sam Altman (OpenAI) and Elon Musk fear a supercomputer that could subjugate us.
Dijkgraaf: “We have to be realistic. Technology is not inherently malicious. We have a responsibility to provide the environment in which these algorithms will decide what their purpose in life is. Because it might be harmless if the goal is to win a game of chess. But very dangerous if the goal is to win a war.
“I see it more as an extra radar screen, with which we will see things that we do not see now. Let’s be careful not to limit ourselves to just two scenarios. Either they’re just stupid robots, or this means the end of the world. There is still a lot in between.”
Do you think AI might help us solve the big problems of our time? To combat global warming or find a cure for Alzheimer’s?
Dijkgraaf: “The biggest and most positive applications of AI are in science. Perhaps AI can help us identify where the opportunities lie. What our scientists should focus on. Because are we worrying about the right problems?
“The great thing is that at many points in history, man has made a leap that came out of nowhere. Something everyone thought was ridiculous and people were going to do it anyway. As a mathematician you learned for a long time that you should never take the square root of -1, because that is not a number. They called it an imaginary number. Well, what a craziness. It now appears that all physical laws make use of precisely these imaginary numbers.”
“There you have it. Humans sometimes also have the ability to do something really crazy. To be able to really make something out of nothing. I hope we will cherish that a little more together.”
A version of this article also appeared in the newspaper on September 7, 2023.