Finding the balance on AI

Rishi Sunak’s global summit on artificial intelligence is a further marker of Britain’s intent to establish itself as a force in the technology revolution. As the capabilities of AI models advance, and their ability to solve complex medical problems and enormously boost productivity, the risk they pose develops also. The purpose of the autumn summit is to devise international rules on how the technology is developed and applied with the goal of minimising this risk.

The great threat in the near term is likely to be the misuse of models by individual or state actors, such as deploying systems to identify and exploit cybersecurity vulnerabilities or producing convincing misinformation at significant scale.

In the future, these systems could become threats of their own. Complex artificial intelligence is hard to control, and we have not fully solved the problem of aligning such models to our interests. There is a risk that firms racing to develop more powerful models may do so without giving sufficient thought to what they may unleash, or that state actors may try to use them to gain an edge over the West.

For the moment, artificial intelligence that exceeds human capabilities in a broad swathe of fields is the preserve of science fiction. But 10 years ago, so was ChatGPT, OpenAI’s revolutionary text-generating chatbot. It is no longer implausible that in the not-too-distant future, we may face powerful AI systems we do not know how to control.

Mr Sunak is right to address this risk before it has occurred. Britain must be at the forefront of the debate on how best to minimise the dangers posed by the most capable new models with potentially dangerous applications.