“They learn how to kill”

GoogleFormer CEO Eric Schmidt has issued a stark reminder of the dangers of artificial intelligence and how vulnerable it is to hacking.

Schmidt, who served as Google’s CEO from 2001 to 2011, warned of “the bad things AI can do,” when asked whether AI was more destructive than nuclear weapons during a fiery talk at Sifted Summit.

“Is there a potential for a spillover problem with AI? Absolutely,” Schmidt said on Wednesday. the Spread risks AI challenges include technology falling into the hands of bad actors, reuse, and misuse.

“There’s evidence that you can take models, closed or open, and you can hack them to remove their guardrails,” Schmidt said. “So, as they’re training, they learn a lot of things. A bad example of this is they learn how to kill someone.”

“All the big companies are making it impossible for those models to answer that question. Good decision. Everyone is doing it. They’re doing it well, and they’re doing it for the right reasons. There’s evidence that it can be reverse-engineered, and there are many other examples of this kind.”

AI systems are vulnerable to attack, using some methods including spot injection and jailbreaking. In a drop injection attack, hackers hide malicious instructions in user input or external data, such as web pages or documents, to trick AI into doing things it’s not supposed to do — like sharing private data or running malicious commands.

Jailbreaking, on the other hand, involves manipulating an AI’s responses so that it ignores its own safety rules and produces restricted or dangerous content.

In 2023, a few months after OpenAI’s ChatGPT was released, users used a “jailbreak” trick to circumvent safety instructions built into the chatbot.

This included creating ChatGPT is an alternate persona called DANwhich is short for “do anything now,” which involves threatening the chatbot with death if it does not comply. The alter ego can provide answers on how to do this Committing illegal activities Or mention the positive qualities of Adolf Hitler.

Schmidt said there is no “good non-proliferation system” yet to help limit the risks of AI.

His comments come amid growing talk about… You have a bubbleas investors pour money into AI-focused companies and valuations appear stretched, drawing comparisons to the collapse of the dot-com bubble in the early 2000s.

However, Schmidt said he doesn’t believe history will repeat itself.

“I don’t think that will happen here, but I’m not a professional investor,” he said.

“What I do know is that people who invest their hard-earned dollars believe that the economic return over a long period of time is tremendous. Why else would they take the risk?”