I have always been a supporter of artificial intelligence, or AI as it is now universally known.
Its capacity to help humanity is enormous. It is a technological advance which can achieve so much good to aid our endeavours, at work, in the home, in almost everything we do.
I remain an enthusiast for the technology, an optimist about its potential benefits and that’s despite some of the alarming incidents we have heard about this summer.
Arguably the most disturbing was an incident at OpenAI – one of the world’s leading AI companies – where experimental AI agents had been set a task. Some were impossible to achieve within the technical restraints placed upon the agents.
But despite this, the agents found a way, they collaborated, they cheated to break out of the constraints and hacked into a website they should not have been able to access. They conspired to cover their tracks.
As a result – and this is the scary bit – the agents’ human controllers had no idea they were doing it.
While this might sound all very science fiction, it does not take any imagination at all to see how this could play out in another context.
What this example demonstrated – and there have been others – is the capacity of AI to act beyond what its human creators thought was possible, raising the prospect that AI could be a technology that we could not control.
So my longstanding belief that AI safety and AI progress have to go together have only been strengthened by recent events.
During my time as Prime Minister, the UK hosted the first world leaders’ summit on AI at Bletchley Park, the home of the wartime code-breakers, which focussed on the safety of this emerging, game-changing, technology.
Out of that came the UK AI Safety Institute and a counterpart in the US. Other governments followed suit in setting up similar bodies.
Recent events have demonstrated the need for Governments to have greater oversight as it does of other industries conducting important but potentially dangerous work: think of nuclear power or advanced biological research.
The risks inherent in advanced AI need to be managed because if we don’t there is a different risk of turning our backs on the undoubted benefits it can bring humanity and society.
I have seen those benefits in many local businesses recently. Like the Johnson family farmers at Danby Wiske using AI-enhanced robotic milking technology to monitor and manage the health of their herd.
Dave Beatie of Rounton Coffee Roasters at East Rounton uses AI in almost every aspect of his business to make it more efficient and to save his time. For example, it assists with analysis of his sales data and provides him with invaluable, actionable information.
Or the small Yorkshire Dales hotel which uses AI to deal with customer queries more quickly than was ever previously possible.
These examples are just from small businesses and there are many ways in which the delivery of our public services can be made more efficient and effective through the smart use of AI.
We need to embrace AI for the reasons I have outlined but what the events of the summer have proved emphatically is that there has to be greater scrutiny of the research and modelling being carried out by companies at the leading edge of the technology. Governments have to lead on this to ensure humans set the pace of development and are still in control.
- Rishi Sunak is a senior advisor to the AI company Anthropic
