Labour party conference: adopting generative AI in government
Policy Director
It can sometimes feel like our politics is stuck in the past, and the same policy debates we always have. But around us the world is changing fast. Artificial intelligence fills some people in Westminster with hope, and some with dread, but it is undeniably part of the economy that is transforming the world, and has an energy and ambition to it that our politics rarely sees. Yesterday, Sam Altman described this period as the start of the Intelligence Age, a fundamental transformation of our global economy based on new ways of organising intelligence.
The challenges of integrating AI into the business of government was the topic of our Labour Party conference panel. The answer? Change is coming whether we like it or not, and it is time for government to get on board.
Minister for AI Feryal Clark, Labour Together’s Kirsty Innes, Co-founder of Extend Ventures Tom Adeyoola and Cohere’s Director of Product for Core Modelling Nick Jakobi joined us to discuss how the State needs to work differently to realise the opportunity of AI to make public services more productive, effective and efficient.
Three key takeaways from me:
- We need to create the conditions for innovation, where the private and public sector can work together. Tom talked about creating “happy playgrounds” in government procurement for start-ups to work with government and thrive. That’s a far cry from most SMEs’ experience of state procurement, where the long timetables and many delays aren’t tolerable to companies with shallower balance sheets, and favour less innovative larger providers.
- AI doesn’t need to be perfect, it needs to be consistently better than humans. We could be talking about failure, bias, risk and the need for human oversight, and be talking about any public service which is provided by humans, not just AI. But sometimes we hold it to an unrealistic standard, and polling shows that the public want autonomous vehicles to be much safer than human drivers. But in some areas of public services we can’t let the final one per cent of safety risk hold us back, because that’s perfectly tolerable. We called for government guidance to clarify this in our recent paper Getting the machine learning.
- Specificity is key. Sometimes the issues of AI adoption in government feel too abstract, like “a solution in search of a problem”. Each speaker had a great example of a practical way they wanted to see AI integrated. My favourite? Saving frontline public servants from doing paperwork, when they could be working with members of the public. For example, Beam’s Magic Notes tool automates large parts of the casework note-taking process. Showing people the benefits of AI in their everyday work will be the fastest way to encourage people to use it.