As the common saying goes, if you're the smartest person in the room, you are in the wrong room. For most of us the reality is that in a few years, there is little chance that we will ever be able to bask in our superiority again. We can leave that to artificial intelligence (AI).
Put in its simplest terms: this year the world changed. Gradually, then suddenly, AI, which for years had been on the cusp of transforming health, education and energy, leapt forward. For those watching closely, yesterday’s release of GPT-4 (the latest version of ChatGPT) was a sign of things to come.
Accompanying OpenAI’s latest product were a raft of innovations from other companies adopting the new technology. The Khan Academy has announced a customised tutor for students and a virtual assistant for teachers, while Duolingo is using it to supercharge its language-learning platform. Danish startup Be My Eyes is using its image-to-text capabilities to help up to 250 million blind or low-vision people identify objects, navigate and much more.
Following Citadel’s recent announcement that it was exploring a licence to help with a whole range of business functions, Morgan Stanley has announced it is rolling out an internal chatbot to unlock the cumulative knowledge of its wealth-management library.
Unsurprisingly, governments have generally been slower to think creatively about its use, with one exception being Iceland. The island nation in the North Atlantic has been worried about the extinction of its language, so is exploring how GPT-4 can preserve it. In doing so, it hopes to also preserve culture and history too.
But more broadly, too little attention has been paid to how these innovations can be used to transform the role of state and provide better service delivery.
Some of the private-sector examples already point to a path forward, not least in education. Morgan Stanley’s chatbot could also easily be applied to policymaking. Bad ideas often refuse to die and get recycled almost ad infinitum in politics as institutional memories are either short or there is an inadequate evidence base to refute them. ChatGPT can help governments stop trying the same thing over and over again – and expecting different results.
The Estonian government has also demonstrated that these innovations can be citizen-facing. Bürokratt, its AI virtual assistant, works on behalf of its citizens, proactively identifying and communicating public entitlements or services they would benefit from while giving them control and transparency over the use of their personal data.
Why the UK Needs a New Mindset
As we also wrote in our recent paper, A New National Purpose, the UK government has to build out its capabilities further. This should include developing and procuring sovereign general-purpose AI systems that can disseminate innovation throughout our public services, as well as learning from visionary labs such as OpenAI and DeepMind to build an elite public-sector research effort in AI. Without it, government will always be playing catch-up to industry.
To do this, the government not only needs to have a mindset shift in its approach to expertise, but also in terms of how it builds the necessary infrastructure. Supercomputing will be critical to this endeavour; to put the challenge in context, OpenAI is using 25 times the total compute capacity available through the entire British state just to train a single model.
Our lack of capacity will restrict our ability to attract talent, commercialise and deploy AI, as well as build global power in a critical, strategic technology. We therefore must invest accordingly.
This will require the UK to deliver a genuinely world-leading supercomputer next year – more powerful than the two-exaflop system the United States will build this year – in order to keep pace with leading AI labs. Second, the current proposal for the UK AI Research Resource to provide at least 3,000 top-spec AI accelerators by the summer of 2023 needs to be upped in ambition by many multiples. The £900 million announced in the budget is therefore welcome, but not sufficient.
Lastly, as other nations race forward, we need to be clear that this is a technology that is essential not only to prosperity, but also to power. Being at the frontier of AI development opens up possibilities of discovering new domains of knowledge and accelerating science and innovation in ways that were previously unimaginable. It is a race liberal democracies must win, but in a way that exercises responsibility.
The UK can demonstrate leadership in this area by building its ambition in multilateral science diplomacy in AI. In leading this effort, it should draw inspiration from initiatives in other areas such as the International Space Station, CERN and ITER. This would crowd in leading expertise and help to embed democratic leadership and values into the frontier of this field.
The last year has completely changed the conversation around AI. Yet despite the noise on the subject, politics is still being too slow to pick up the signals. The adoption curve is going to happen fast and some of the most exciting innovations are happening in this space. Government needs to grasp the opportunities.