As Open AI’s chief scientist, Ilya Sutskever worries about the power of his own invention. Now he regrets his involvement in the expulsion of Sam Altman – why remains unclear.
«Artificial intelligence will solve all the problems we have today. It will solve problems in the labor market, it will cure diseases, it will fight poverty. But it will create new problems. Artificial intelligence has the potential to create everlasting dictatorships.”
With these words the Canadian computer scientist Ilya Sutskever warns in one Dokumentation des «Guardian» from the risks of artificial intelligence. According to Sutskever, AI will one day be so powerful that its relationship with humans will be like that of humans with animals: “We like animals. But when it comes to building a highway between two cities, we don’t ask the animals for permission.”
Morality or profit – what is more important?
Sutskever himself is one of the founders of one of the most influential AI companies: Open AI, the company behind Chat-GPT. Sutskever sits on the board of directors at Open AI and is considered the driving force behind the dismissal of CEO Sam Altman, which was surprisingly announced on Friday. The move caused great dismay among employees – on Monday Sutskever finally announced that he regretted the decision. He didn’t want to harm Open AI. He also signed a letter demanding the resignation of the board of directors.
I deeply regret my participation in the board’s actions. I never intended to harm OpenAI. I love everything we’ve built together and I will do everything I can to reunite the company.
— Ilya Sutskever (@ilyasut) November 20, 2023
It’s the latest chaotic turn in a battle over morality and profit pitting Altman and Sutskever against one another. For the latter, CEO Altman’s plans went too far. If you don’t take the risks of AI seriously, says Sutskever, one day it could become too powerful. He himself is one of those who made the development of language models like Chat-GPT possible in the first place.
Sutskever, born in Soviet Russia in 1984 and raised in Israel, studied at the University of Toronto under Geoffrey Hinton, who is considered the “godfather of artificial intelligence.” Around a decade ago, he and Hinton developed the neural network Alexnet, which used deep learning to analyze thousands of photos and recognize objects such as flowers or cars. The associated company was soon bought by Google for $44 million, where Sutskever worked from then on.
Elon Musk brought Sutskever to Open AI
Another AI enthusiast at Google soon became aware of Sutskever: Elon Musk, who regularly warns about the potential risks of artificial intelligence. He convinced the young computer scientist that Google founder Larry Page wasn’t giving enough thought to these security risks. Sutskever should therefore join the Open AI founding team around Musk, Sam Altman and Greg Brockman.
At Open AI, which was founded as a non-profit organization in 2015, Ilya Sutskever became Chief Scientist. As such, he played a central role in the development of the Chat-GPT language model and the Dall-E image generator. With great success: When the chatbot Chat-GPT 3 was made available to the public almost a year ago, it attracted more than a million users within just a few days.
But the question of how to deal with potential risks of AI also became more and more of an issue in Open AI.
Four years after it was founded, Microsoft comes along
Elon Musk left the company’s board of directors in 2018 after a power struggle between him and Sam Altman. Earlier this year, Musk wrote in a tweet that Open AI had become a closed, profit-driven company under the control of Microsoft.
OpenAI was created as an open source (which is why I named it “Open” AI), non-profit company to serve as a counterweight to Google, but now it has become a closed source, maximum-profit company effectively controlled by Microsoft.
Not what I intended at all.
— Elon Musk (@elonmusk) February 17, 2023
In fact, after Musk’s departure, Open AI increasingly turned away from its original non-profit goal. In order to have more money for developing products, Sam Altman founded a profit-oriented branch of the company in 2019, which, among other things, attracted billions in investment from Microsoft.
Conflict between Altman and Sutskever comes to a head
However, control of the company remained with the non-profit board of directors – and therefore also with Ilya Sutskever. The ethical questions raised by the rapid success of the language models led to increasing conflicts between Sutskever and Altman. According to anonymous sources quoted by the Bloomberg news agency, their opinions differed on how quickly the generative AI products should be commercialized and what steps are necessary to minimize potential damage.
While Sam Altman, as the face of Chat-GPT, has spent a lot of time over the past few months talking to politicians about the opportunities of AI and easing their concerns (“The advantages of our technologies clearly outweigh the disadvantages”), those of Ilya Sutskever continued to grow.
Superintelligences as a threat
Sutskever is particularly concerned about the possibility that a type of artificial superintelligence that would be superior to humans could get out of control. In July, Open AI launched a new project led by Sutskever that is dedicated to technical solutions for better control of artificial intelligence. Artificial superintelligence, the company said in the announcement, “could be very dangerous and lead to the disempowerment or even extinction of humanity.” You have to ensure that the computer pursues the same goals as humans.
Sutskever is not alone with these worries: in March of this year More than 1,300 tech entrepreneurs, including Elon Musk, published a letter in which they called for a pause in the development of artificial intelligence. And Sutskever’s doctoral supervisor Geoffrey Hinton announced his departure from Google in May, to be able to speak freely about the dangers of AI from now on.
Sutskever regrets his actions
Sutskever is also aiming for a break in AI development. In the future, he wants to shift his focus from developing new language models to preventing an overwhelming superintelligence. Sam Altman, on the other hand, will move to Microsoft, as it was announced on Monday, to lead a new research team together with Open AI co-founder Greg Brockman.
More than 700 of Open AI’s 770 employees announced Monday that they would join Altman and Brockman in leaving the company. At the same time, the U-turn at Sutskever followed. He said he wanted to do everything he could to bring the company back together.
Whether he will succeed in this is unclear – as is how the development of artificial intelligence will continue. For Sutskever, however, one thing is certain: “It will be monumental, earth-shattering. There will be a before and an after. »
The post first appeared on www.nzz.ch