UvA scientists sign open letter: “Stop the development of AI systems”
The development of AI systems more powerful than the current ChatGPT-4 should be put on hold for the next six months. This is stated in an open letter from the non-profit institute Future of Life, signed by UvA scientists alongside big names such as Elon Musk including researcher Media Studies Amir Vudka. “GPT-5 could be a game-changer.”
Amir, why did you sign the letter?
“First of all, the letter itself won’t accomplish much. I don't think the letter can influence the big AI companies, as there is far too much money involved for that. Still, I signed the letter because it is crucial to think about the dangers of AI.”
Isn't it questionable that signatory Elon Musk himself founded the company OpenAI that “trained” ChatGPT-4?
“I think it shows he knows what he’s talking about. And he is not the only one who helped train this system who is now sounding the alarm. In addition, of course, he may have a financial motivation.”
What is at stake?
“In the worst-case scenario, AI will pose an existential threat to humanity. Last week, a team of AI scientists that trained ChatGPT-4 published a paper describing how GPT-4 shows ‘sparks’ of artificial general intelligence (AGI). That's a huge leap in development. AGI is a form of AI that, like humans, can learn entirely on its own, without the need for specific commands. You can quickly lose control over such a system. Earlier it was doubted that we would reach that form of artificial intelligence, and if so, perhaps not for another 20 years. Now there is an indication that the next AI model is already showing these traits. GTP-5 could become a game-changer. In addition, AI could cause a lot of disruption in society. The prediction is that in five years, 25 percent of jobs will have been replaced by AI systems. In fact, I think 70 percent of jobs will be replaced in the next 10 years.”
What can be accomplished in a six-month pause?
“The most important thing is to align AI with people’s needs so that it works for us and not against us. And we need to think about how we introduce this radically disruptive technology into society. That’s what universities, funders, companies, and government need to do together.”
Is the topic very much alive at the UvA?
“The discussion at UvA is mainly about what impact ChatGPT has on education. Students can do assignments and take exams with ChatGTP but cannot be checked for plagiarism. Colleagues researching artificial intelligence worry about bigger issues. Not all of them; I don’t want to paint a doomsday scenario, either. But we have to be careful.”
UvA’s associate professor of theoretical physics Christoph Weniger, professor emeritus of Science & Technology Studies Stuart S. Blume and physicist Jan Pieter van der Schaar signed the open letter.
