In a candid interview and a Q&A session with experts, he emphasizes that the world is not taking the alarm signals seriously.
Sam Altman, the head of OpenAI and one of the most influential voices in the global debate on Artificial Intelligence, has warned of three categories of risks that he considers the most frightening regarding the future of this technology. In a frank account and in a question-and-answer session with people in the field, he emphasizes that the world is not taking the alarm signals seriously and that society could face serious consequences, both from the use of AI for malicious purposes, as well as from the gradual loss of human control over it. Three scenarios, more specifically, are what scare Sam Altman:
WHEN THE “BAD ONE” GETS SUPER-INTELLIGENCE FIRST
The first is when a bad person gets the superintelligence first and uses it badly before the rest of the world has a version powerful enough to defend itself. So an enemy of the United States says they’re going to use this superintelligence to design a biological weapon to destroy the American power grid, to infiltrate the financial system, and to take everyone’s money. That’s the first category. And I think the biological capability of these models, the cyber capability of these models, is getting very large. We keep raising the alarm bells about this. I think the world is not taking us seriously. I don’t know what else we can do here, but this is a very big thing that’s coming.
LOSS OF CONTROL OVER SYSTEMS
The second category is what’s commonly called loss-of-control incidents, where it happens a lot like in science fiction movies. The AI says, ‘I don’t want you to turn me off. I’m scared, I can’t do this.’ You know, what. There’s a lot of work that we and other companies are doing on adapting models to prevent this. But as these systems become so powerful, this remains a real concern.
ACCIDENTAL TAKING OF POWER BY HIM
The third one is harder for me to imagine, but quite scary. I’ll explain what it is and then I’ll give a short-term example and a long-term example. This is the category where models just take over by accident. They never wake up, they never do things like in science fiction movies, they don’t open forbidden doors, but they become so ingrained in society and so much smarter than us that we can’t really understand what they’re doing, but we’re forced to rely on them. And even without a single bit of malice from anyone, society can veer in a strange direction. This phenomenon, I think, is a very big deal.
In the short term, you can see when people maybe, we call this excessive emotional support, rely too much on ChatGPT.
There are young people who say, ‘I can’t make any decisions in my life without telling ChatGPT everything that’s going on. It knows me, it knows my friends, I’ll do whatever it tells me.’ That to me is very bad. The long-term example is: What if AI gets so smart that the president of the United States can’t make better decisions than ChatGPT 7 recommends, for example? It might be the right decision in any given case. But it means that society, collectively, has transferred a significant part of decision-making to a very powerful system that is learning from us, improving from us, evolving with us, but in ways that we don’t fully understand. So that’s the third category of ways things can go wrong.
