Sam Altman warns: AI will also cause ‘bad things’

According to OpenAI CEO Sam Altman, the rapid development of artificial intelligence is inevitably accompanied by negative consequences. Nevertheless, he believes that society should be willing to accept some of those risks, because the benefits of AI can be much greater, according to him.

In an interview with Politico's Decoded podcast, Altman argues that a world without abuse, fraud or major cyberattacks by AI is not a realistic starting point. At the same time, he does draw a clear line in scenarios in which people actually lose control of artificial intelligence.

Altman made his striking statements during a conversation about how AI companies and governments should deal with the risks of the technology.

According to the OpenAI CEO, the world must accept that “bad things” will happen as a result of AI. According to him, there are enormous social advantages to this.

“I wouldn't choose a situation where we make sure there are no major hacks, no abuse of this technology, zero scams, and none of the other bad things that will happen,” Altman said.

According to him, people would use AI much more often for positive than for harmful purposes.

"I think that people will do an enormous amount, orders of magnitude more, good things than bad things," says the OpenAI CEO.

Altman is therefore opting for a relatively light approach to regulation. According to him, the development and use of AI should not be limited so much that possible benefits of the technology are lost.

That attitude plays an important role in the discussion about the future of AI. As models become more powerful, so does the fear that they can be used for cyberattacks, fraud, and other forms of abuse.

However, Altman does not want to accept every risk. During the same interview, he made it clear that, according to him, there is a limit to truly catastrophic scenarios.

According to him, this includes a "serious loss of control over AI". In such a situation, an AI system could behave in such a way that people can no longer intervene effectively.

During the interview, it was not clear where exactly the line lies between acceptable damage and an unacceptable risk.

This distinction is becoming increasingly important as AI systems are able to perform tasks more independently. As a result, the discussion is no longer just about what a chatbot says to a user. The question of which actions AI may perform independently is also receiving more and more attention.

Despite his advocacy for relatively light regulation, Altman doesn't think AI companies should go wild when their technology causes harm.

He advocates a legal framework that clearly states who is responsible when something goes wrong.

“If something goes wrong with our models during training, there will have to be a form in which we are responsible for it,” Altman said.

According to him, American policymakers should continue to debate this. In doing so, he chooses a combination of space for rapid technological development and liability when companies make mistakes.

The discussion about that balance becomes more urgent as AI systems become more powerful. Altman seems to make one starting point clear: according to him, complete safety is not a realistic condition for further development, but some risks are too great to accept.