One of the pioneers of artificial intelligence has warned that governments are not protecting against the dangers posed by future superintelligent machines.
Professor Stuart Russell told The Times that ministers supported brushing aside the burgeoning AI industry despite warnings that civil servants could create an existential threat. .
A former Downing Street and White House adviser, Professor Russell is co-author of one of the most widely used AI textbooks and lectures on computer science at the University of California, Berkeley.
He told The Times that a ChatGPT-like system that can pass exams and produce prose could form part of an uncontrollable superintelligent machine.
“How can I permanently maintain power over beings more powerful than myself?” he asked. “If you don’t have the answer, stop researching. It’s that simple.
“The risks cannot be higher. If we do not control our civilization, we will have no say in whether we will continue to exist.”
In March, he co-signed an open letter with Elon Musk and Apple co-founder Steve Wozniak warning of an “uncontrolled race” going on in the AI lab.
The letter warned that the Institute was developing “an ever more powerful digital mind that no one, not even its creators, could understand, predict, or reliably control.”
Professor Russell is involved in the development of a system to monitor the Nuclear Test Ban Treaty at the United Nations and was asked to work with the government earlier this year.
“The Ministry of Foreign Affairs … has spoken with many people and has come to the conclusion that the loss of control is plausible and has very serious consequences,” he said.
“And the government has launched a regulatory approach that says, ‘There’s nothing to see here…we welcome the AI industry as if we were talking about manufacturing cars or something like that.'”
He said it would take “time we might not have” to make changes to the AI’s technical underpinnings to add the necessary security measures.
“I think we did something wrong at the beginning. he said.
“Unless the sole purpose is to benefit humans, you are actually creating a competitor, which is clearly stupid.
“We don’t want a system that mimics human behavior… basically we have a human-like goal and we train the system to pursue that goal.
“I can only imagine how disastrous it would be if we had a truly capable system to pursue such a goal.”
He said there were signs that politicians were recognizing the risks.
“We kind of get the message and we’re kind of stuck on what to do,” he said. “This is how I feel now.”
The government has set up an AI-based models task force to “lay the groundwork for the safe use of underlying models across the economy and ensure the UK is at the forefront of this crucial AI technology”.
