What can AI do for the public sector? What can AI do for the public sector? There has been a lot of hype about that. A lot of hopes, a lot of dreams, and a lot of fears. But I think if you try to look at it in a pragmatic way, that really makes sense. So there are probably four things that AI can help government to do. And they're important things, they're some generic tasks which might be applied to any policy sector. One of them is measurement and detection. There's all kinds of places where we need to sort of understand, for example, whether some harmful behaviour of some kind in an online environment, for example, or where there's maybe a failing school or hospital or a factory that needs inspecting. AI is good for that, for detecting irregularities in data and potentially being able to measure them. Another thing that AI is really good for is simulation. How can we kind of create a kind of twin of the real world, if you like, so that we can test out policy interventions before we put them into practice? During the pandemic, for example, there was a huge furlough scheme, a big system of loans and a kind of stimulus package for companies trying to get out of the pandemic. Again, AI is good for being able to try those things out and explore any unintended consequences without having to suffer them. And then there's also the question of prediction. How can you predict how many school places you're going to need or how many hospital beds you're going to need or the next economic crisis or the next pandemic, God forbid? So that's something that government needs to do. It's something that government has traditionally not been very good at, and AI can really help with that as well. And the final point I'd like to make is actually half negative, half positive. There's a lot of ethical dilemmas that government has to face, and these technologies can kind of highlight those ethical dilemmas. So if you take something like predictive policing, using prediction in policing or criminal justice more generally, the problem with that is the kind of decisions or the kind of insight that technology gives can be biased because it's based on biased data. And that biased data is really revealing bias in human decisions that have been made in earlier years. And what's exciting about the technology is it reveals those problems and it forces us to think about them and to think about what we can do about them and how we can try and counter them. So that's another important element of AI development.