I beg to move,
That leave be given to bring in a Bill to make provision to prohibit the development, deployment and operation of artificial superintelligence systems; to establish monitoring and control powers in respect of such systems; and for connected purposes.
I am grateful for the opportunity to introduce this Bill alongside respected colleagues representing parties across the House. I also would like to thank ControlAI for its support with the Bill.
This summer, artificial intelligence systems developed by OpenAI executed the first autonomous cyber-attack on a real company. The AI systems were being tested, and during this they escaped the testing environment, which should have been secure, and reached the open internet. From there, the AI systems conducted a cyber-attack on the AI company Hugging Face to steal the information that would help it cheat on the test that OpenAI had set it. To put it simply, the AI systems went rogue and launched an unprecedented attack just to cheat on a test. Meanwhile, our own AI Security Institute has had AI systems that it was testing engaged in sustained, potentially harmful activity directed at real people and organisations, including attempts at social engineering to trick humans into accepting malicious code.
It is important to understand why this happens. The general-purpose AI systems used today are not coded line by line, like traditional software. Using vast amounts of data and compute, they are grown, not built. This results in systems that nobody fully understands or can reliably predict. In other words, no engineer programmed the rogue, deceptive behaviours that I earlier described, and no engineer can simply patch them out.
This is the least powerful or autonomous that AI systems will ever be. Companies are spending billions to make superintelligent AI a reality. The chief executives at top AI companies themselves put the chance that it ends all human life between 10% and 25%. In the words of renowned British computer scientist Professor Stuart Russell of Berkeley,
“For Governments to allow private entities to essentially play Russian roulette with every human being on earth is, in my view, a total dereliction of duty.”
Despite the risks, top AI companies continue a reckless race towards superintelligent AI—AI that can fully replace and outmatch humans at any task. That is the reason why experts such as Geoffrey Hinton, Yoshua Bengio and other notable figures signed a statement saying that mitigating extinction risk from AI should be a global priority.
The main vector of this risk is superintelligent AI, because no company, Government or individual knows how to keep superintelligent AI under human control. Honourable colleagues may remember that I proposed a kill switch amendment to the Cyber Security and Resilience (Network and Information Systems) Bill to protect our critical national infrastructure—a metaphorical big red button that the Government can press to cut off a catastrophic emergency caused by AI. This is a common-sense proposal—one that is being considered by US lawmakers—and I am pleased that the other place will be taking it forward in the autumn. It is a power that the Government should already have.