London Tech Week - PM Rishi Sunak promises to regulate AI, but admits UK not up to speed in spreading the benefits to all

Sunak on AI

London Tech Week kicked off today with British Prime Minister Rishi Sunak debating AI regulation and ethics, even if the mainstream media headlines were inevitably more focused on his response to the latest political shenanigans by his predecessor Boris Johnson.

Leaving such party games to one side, Sunak keynoted at Tech Week with a short speech, heavily focused on the UK being “an island of innovation”, but one where the “tectonic plates of technology are shifting” to create more urgency for action. And it was, inevitably, on AI that he placed most emphasis:

If our goal is to make this country the best place in the world for tech AI is surely one of the greatest opportunities before us.  As Chancellor, I doubled the number of AI scholarships because even back then I recognised the potential of AI as a general-purpose technology.

But, there’s a but:

The possibilities are extraordinary. But we must – and we will – do it safely.  I know people are concerned. The very pioneers of AI are warning us about the ways these technologies could undermine our values and freedoms through to the most extreme risks of all.  And that’s why leading on AI also means leading on AI safety.

AI doesn’t respect national boundaries, he noted, so global co-operation is needed around regulations and ethics, hence the UK hosting a summit on global AI safety later this year:

This is our strategy for safe AI - to lead at home; to lead overseas; and to lead change in our public services. All part of how we meet our goal of making this the best country in the world for tech. 

So far, so pretty much standard political pitch for AI and the wider topic of making Brexit Britain a place for inward investment and a global tech hub. But it was when Sunak then sat down for a conversation with Demis Hassabis, CEO and co-founder of Google DeepMind, and other business leaders that the discussion became more engaging, as he drilled down further on the UK attitude to regulation:

There's always, when it comes to technology and new businesses, a balance to be struck between regulation that supports innovation, but puts appropriate protection in place.  I just think the UK has a track record of getting that balance right. Not everyone does.

Salesforce UK and Ireland CEO Zahra Bahrololoumi raised the importance of the pace and scale of international collaboration, citing the COVID vaccine efforts as a case in point of how much this matters in practice. The challenge, she pointed out, is to have a regulatory framework that can drive innovation, but at the same time balance changing regulatory requirements globally.

Sunak’s response was guarded:

It's not a formula, right...what I touched on is getting this balance right between protection and safety and guardrails, but not throwing the baby out with the bathwater, and making sure that we continue to promote innovation. Now, I said it, it's not a formula; it's just a way that we approach these things.

I think it comes from the equity [that] stretches all the way back to The Enlightenment, that led the Industrial Revolution. Our approach to these things is probably just in our cultural DNA, which is a great asset for the UK. I think we approach things with principles. First, logic, which I think is right and I think that really helps us get the regulation. I think we're big on engagement...You've got a government that is focused on engaging with industry to make sure that we understand how the innovation is happening, to make sure we're creating a safe space for it to happen.

Regulatory sandboxes are a good example of such spaces, he suggested, pointing to the FinTech sector:

The FCA [Financial Conduct Authority] was the first regulator in the world to create a regulatory sandbox that's now becoming a scale box that's been exported around the world. So I think that we've got a good track record in doing that.

But there are genuine concerns that people have, he went on:

One is the socio-economic impact, about the impact on people's jobs. And then there are obviously risks around misuse. How do people use AI to do things that are bad?  People have always used technology to do bad things, this is not a new thing. There's what we call 'tool use' and AI is a capability that provides up new opportunities for malicious actors.

Then there's what I guess technologists would call 'capability overhang', which speaks to the last category of risks where you start to trend into a space where people are talking about very significant risks that some would call existential, others would say alongside things like pandemics or nuclear war, and that's obviously concerning. That's why I think it's important that we put guardrails in place to make sure that we develop and exploit this technology in a way that's safe and that is secure. That's why the Summit is important because it will bring together people to consider those risks, make sure that we appropriately plan for them, discuss the right guardrails to put in place and the research and evaluation that we're going to need to do to make sure that we're on top of it.

These comments were met with approval by DeepMind’s Hassabis:

There are attendant risks with any new transformative technology and AI is no different in that respect. I agree with the Prime Minister's different buckets or categories of risk, and I think they need to be mitigated in different ways, from bad actors using these technologies in bad ways, from access issues to near term risks of things like bias or deep fakes. There are new technologies along the way, like encrypted watermarking, which will help with that. And then there will be sort of more long-term technical AI risks, things like alignment, controllability, interpretability of these systems.

There's uncertainty around those risks, and we need to understand and research those systems a lot more over the next few years to better have a better handle on what the boundaries are of the systems, including the risks, and then we can put the right guardrails in place. Government can put the right guardrails in place. In a situation where there's a high degree of uncertainty, but huge potential impact either way, I think the only the right way to proceed is with the precautionary principle. You know, be proceed with exceptional care, be optimistic about what we can do with the opportunities and, and  things like the scientific methods to study and carefully analyze these systems as they get increasingly more powerful in the future.

My take

Alongside the important comments around regulation and ethics, there was one other topic that Sunak raised and responded to with some candor. This was the need to make sure that the incentives and the environment are right to diffuse AI technology across the economy, an area he admitted where there was work to be done:

It's no point if we create these things and then they don't get spread, so the benefits are not realised fully. I think that's actually something where we probably need to do a better job. There's actually good academic research that the kind of diffusion of technology and best practice across the economy in the UK is probably a bit less than is optimal, or less than other countries.

So I think one thing we all need to keep an eye on is not just the incredibly switched-on innovative people in this room [at London Tech Week]. We've got to figure out how we spread that mindset to every small and medium sized company everywhere in the country, because that's how you really get the benefits of these things realized.

Sunak is quite correct on this point. That said, he might then wish to contemplate how much such incentive is flourishing under the macro-economic policies enacted by the current administration at the UK Treasury? Government is working at pace, Sunak insisted in response to a question from Erin Platz, CEO of HSBC Innovation Banking, on what to expect in terms of scaling up the AI sector:

We are not in the business of resting on our laurels. I think we all get that this is moving at a pace that means we [have to] move at that same pace. You've seen that in the creation of [the UK Government's] AI Taskforce [that was] announced a while ago, and that will be modelled on the Vaccine Taskforce. It's going to have the agility and flexible and delegated authority to do what it needs to do to build a leadership position in in AI. That [Vaccine Taskforce was] the Government doing things differently, and acting a little bit more like all of you in this room. That's very much come from my own experience, and career. I want to bring that same way of thinking to government as well.

The ambition cannot be argued with. The delivery and execution on the other hand remains to be proven in practice. That said, Sunak was more impressive in his Q&A than I would have given him credit for - not clinging to same four or five parroted PR-doctored talking points was an improvement - and he clearly does have a passion here. Whether that translates into the benefits for all that he talked about is another matter.