Ed Bithell

Strategy and partnerships at Fractile

Also writing on trade, tech and industrial policy

Former UK civil servant and diplomat, 好異史者, he/him

(usual caveats)

[Edward Gibbon @EdwardEGibbon]

If London had Washington DC’s homicide rate (2024), it would have around 2,500 murders a year. The true figure is about 100.

[Ed Bithell @Ed_Bithell]

Advancing AI brings security concerns alongside economic opportunities, and serious state capacity is essential. Building on the success of @AISecurityInst the government should go further to bring in the best possible people to work on the critical issues facing us today

[Ed Bithell @Ed_Bithell]

Thanks to @siobhankennedy4 for speaking yesterday about why now is the crucial moment for the UK government to ensure the UK is at the forefront of global AI and tech progress - it's a hugely exciting opportunity and we have to seize it!

@Channel4News

[Ed Bithell @Ed_Bithell]

AI is advancing at a phenomenal pace, and the UK is very well placed to be an indispensable part of the value chain. The UK should double down on our competitive advantage in domains that require a world-class talent, R&D and business environment, like chips

[Ed Bithell @Ed_Bithell]

At DSIT, Kanishka has been a v effective champion of bringing in top talent to tackle the most important challenges, in excellent teams like @AISecurityInst and @UKSovereignAI. Keeping this focus on UK state capacity is vital to seizing the opportunity we have right now in AI.

[Ed Bithell @Ed_Bithell]

Great to speak to @siftedeu about the opportunity @kanishkanarayan has to make an even bigger impact in his new Cabinet role.

Machinery of government changes grab headlines, but the factors that really drive long term impact are people, and clarity of purpose.

[Ed Bithell @Ed_Bithell]

@steamedhamms I have no idea whether it's the same experience but I'm gonna find out

[Ed Bithell @Ed_Bithell]

@steamedhamms There's an IMAX at the Science Museum!

[Tom Westgarth @Tom_Westgarth15]

Great to speak with @tim about what it might take for the new DSIT MoG changes to work. Feels like a fork in the road moment for state capacity in technology! Some views from me in this well-researched piece, plus expanding a bit below:

1. Give the Cabinet Office AI Minister real authority to drive through the most ambitious work.

2. There is also a fantastic chance for @jreynoldsMP to immediately signal that he is seizing the opportunities of our tech future, and catalyse the confidence of a sector driving growth and investment in the country.

3. Lots of focus right now on structure, but people will matter above all else. What DSIT managed to do when it comes to attracting exceptional talent across research, AI, and entrepreneurship is exceptionally rare. Cabinet Office and DBIST should internalise what has been done here, and spread those best practices to make their departments purr!

4.) Spreading the AI brief across multiple departments will allow for some important areas to be prioritised, but it also strengthens the need for an effective No 10 AI operation, that can adjudicate between departments that end up in disagreement with one another. No 10 have assembled a great AI unit and tech team to address that, and their work continuing will be especially important.

[Ed Bithell @Ed_Bithell]

Really positive development - @KanishkaNarayan is fantastically capable and credible on AI.

Key question is indeed what levers he will have in his new role, as well as whether HMG keep excellent teams in AISI, No 10 etc, well-empowered in the new structure

[Joe Bambridge @JoeBambridge1]

SCOOP: Burnham to create an AI Taskforce in CO and the PM's office, overseen by Kanishak Narayan. AISI also moving to CO/No. 10.

Plenty of outstanding questions, like whether taskforce will have levers over stuff like copyright, SovAI, compute etc

https://t.co/J4pNzM8nzA

[Matt Clifford @matthewclifford]

Fantastic to have Kanishka attending Cabinet as AI minister. He understands the scale and urgency of the opportunity and has brought great energy and clarity to the role over the last year.

[Ed Bithell @Ed_Bithell]

@alexschristian They need to make him one that isn't a glove, he made more tackles than saves in that game

[Tom Westgarth @Tom_Westgarth15]

SIX IDEAS FOR ANDY BURNHAM’S SOVEREIGN AI POLICY

The moment Andy Burnham walks into Number 10, the briefings will start. The new Prime Minister will learn about the nuclear deterrent, about state and non-state threats. Then the securocrats will tell the Prime Minister about the critical importance of AI to our security and prosperity.

Dylan Rogers (ex @AISecurityInst) and myself (ex @UKSovereignAI) were chatting recently about the lessons from working on AI policy under Keir Starmer. Lots of good work underway. Some progress. So much more to be done.

AI might not be a priority for Burnham and his closest allies today. When so many of the basics are broken, it is easy to understand why. But events will soon force Burnham to confront the technology of our time. Here are six suggestions for how AB should approach AI. This is also posted on Dylan’s Substack (link below).

1. DON’T SETTLE FOR SECOND BEST AI SYSTEMS

The UK needs access to frontier AI. In many domains, marginal differences in capability matter. Take cybersecurity and scientific research as examples.

AISI research describes models like Mythos as a ‘step up’ in cybersecurity capabilities. On expert level tasks, which no model could complete before April 2025, Mythos succeeds 73% of the time. In this context, relying on sub-frontier systems means accepting vulnerability.

Scientific research similarly privileges frontier AI. Creating hypotheses, running experiments, and evaluating data requires systems that reason over long time horizons. Small differences in per-step reliability determine whether an autonomous research loop closes at all. Relying on sub-frontier systems could mean ceding the UK’s edge in scientific discovery.

In two of the areas the new Prime Minister plans to differentiate himself from his predecessor, defence and innovation, access to frontier AI is essential.

2. SEEK LEVERAGE THROUGH INTERDEPENDENCE

There is justifiable interest in building frontier AI here at home. But the UK will not win by playing catch up with the US.

Well-capitalised efforts in France (Mistral) and Canada (Cohere) demonstrate the difficulty of closing the gap to the US. As new compute infrastructure comes online, and the leading US companies begin to automate aspects of AI research itself, the gap will only grow.

There is certainly value in a fallback option for the UK. But we should acknowledge the reality that a sovereign LLM is a hedge, not a route to the frontier.

When we bet on AI companies, it should be labs pursuing novel research directions which capitalise on the limitations of LLMs. Following the same recipe five years late only makes sense if the state plans to backstop a second best system.

So instead of competing with the leading US companies, the UK should seek leverage through interdependence.

This leverage can only be arrived at by building infrastructure to enable frontier AI deployment, creating conditions for category defining companies to emerge, and adopting a new approach to how we leverage scale-up finance for the very best firms.

3. BUILD AI INFRASTRUCTURE THAT BENEFITS COMMUNITIES

Building infrastructure is one of the best ways to create leverage over the US. Compute is likely to be supply constrained in the near term. This makes even small buildouts matter.

Many have been critical of the government’s AI Growth Zones, claiming they are creating value for US companies with little captured by the UK. The Prime Minister should establish a new social contract for compute infrastructure: one that rapidly delivers the benefits of AI to the public.

Currently, the costs of building datacentres are borne by local communities, but the benefits are more distributed. This needs to change for durable democratic consent for infrastructure that is absolutely essential to the UK economy and our global position.

This could involve letting local authorities keep business rates from datacenter builds, negotiated on the ground, not in Whitehall, or guaranteeing consumers will not see higher energy bills. In return, the government should want infrastructure built fast.

4. CREATE THE CONDITIONS FOR CATEGORY DEFINING COMPANIES

Beyond infrastructure, it will be difficult to identify assets that confer leverage over in advance. Even the most expert state would have struggled to predict toilet maker Toto as a winner from the AI infrastructure buildout.

This does not mean giving up. Instead, it advantages ecosystem-level interventions. Ultimately, sovereign AI policy should be startup policy.

The new Prime Minister should create the conditions for category-defining companies to start and scale in the UK. These companies will provide both leverage (including over frontier US AI labs) and economic opportunity. Building these companies therefore represents both a unique economic life raft as well as an existential challenge to the future of the country.

There are many steps the government could take to improve the environment for startups. Talent is evenly distributed across the UK, but the ability to start companies is not shared. We need high-speed rail to connect our second cities, and realise agglomeration effects. We should continue to expand EMI, and eliminate non-competes across most sectors.

While enabling the best British entrepreneurs, we should also lower the barriers to global talent. This could include advertising visa routes more broadly, reducing application costs and waiving the NHS surcharge, and starting proactive headhunting programs. Why not offer junior IMO medallists a university place here in the UK?

5. CONSOLIDATE SCALE-UP FUNDING, UNDER THE RIGHT LEADERSHIP

What is being built at the Sovereign AI Fund (SovAI) has the potential to be the defining intervention in the UK’s industrial strategy. But its budget over the Spending Review is the same as the government’s annual pothole budget. It also is focused on smaller cheques for earlier stage companies.

The maximally ambitious version would combine the deepening expertise, network and speed of SovAI with the scale of the British Business Bank (BBB), enabling £2bn annually to be spent on helping to scale the most ambitious UK companies. This could be called British Sovereign Capital, as advocated by Andrew Bennett. This wouldn’t cost any additional money.

This would only succeed with a truly world class CEO leading British Sovereign Capital. The current BBB CEO is due to step down in September. Ensuring the appointment of an exceptional new CEO with an extensive understanding of frontier AI will be decisive, and gives the new Prime Minister an opportunity to make a bold statement of ambition.

6. KEEP DSIT AS A DISTINCT DEPARTMENT

Responding to the scale of AI’s challenge and opportunity demands a department that is dynamic and has focus. Many of the successes of recent Labour and conservative governments, such as the AI Security Institute, only could have emerged through a department like DSIT and with senior No10 backing. Messing around a lot with its functions, or merging it with another department like DCMS or DBT, would be a major mistep. The stakes are too high with AI to be slowed down by a machinery of government change that will dilute focus. A Cabinet Minister for AI would be a great move, but ensure that the appointee understands the technology and is happy to stay in the role for the rest of their parliamentary term, rather than just looking for the next promotion.

CONCLUSION

Not all of the problems this country faces relate to Artificial Intelligence. But how the country responds to this new technological era will determine whether we have the agency to address our most fundamental challenges. Burnham should back the builders. Develop a new social contract for AI infrastructure. Upgrade local transport. Don’t shutter DSIT. This is the way to reindustrialise the country, give the country their first pay rise in a generation, and to revive our sense of purpose and destiny.

[Ed Bithell @Ed_Bithell]

@lewis_goodall Finally on the workforce capability gap, I totally agree we must give more latitude to get the best people in. However, DSIT (esp AISI) have successfully innovated here under Sunak and Starmer. I would fear a retrenchment of CS standard practice if folded into DBT

[Ed Bithell @Ed_Bithell]

@lewis_goodall No doubt it's possible to keep things at the top of the agenda without a department, e.g. if the PM is extremely engaged on the issue and drives it personally through all departments. But requires consistently high energy engagement at the top, and budgets etc less protected

[Alicia E. Stallings @ae_stallings]

(Occurs to me that what Classical Greeks would hate most about the Odyssey casting, would be that the women aren't played by men.)

[Ed Bithell @Ed_Bithell]

@lewis_goodall That last point also applies to the sheer volume of big urgent decisions DSIT ministers are making - would be concerned for those decisions to be folded into the box of a Sec of State also responsible for many other major issues like steel, FTAs, etc

[Ed Bithell @Ed_Bithell]

@lewis_goodall Also concerned by the prospect of weeks if not months of the inevitable friction from organisational shifts, moving buildings, engaging and disengaging IT systems, etc, at a time when AI and other parts of the tech landscape are moving at incredible speed