Three Big Ideas #11

Following France on nuclear, future social science frontiers, and a ‘shock UK regulatory coup’

Welcome to our weekly Three Big Ideas roundup, in which we serve up a curated selection of ideas (and our takes on them) in entrepreneurship, innovation, science and technology, handpicked by the team – and today, with a guest contribution!

---

⚛️ Jeremy Driver, Britain Remade

Britain used to be a nuclear superpower. The atom was first split in Britain. In 1956, Britain opened the world’s first full-scale commercial nuclear reactor. Less than ten years later, it had built 21 more. As late as 1965, Britain had more nuclear reactors than the rest of the world combined.

Yet Britain hasn’t completed a new nuclear power station in almost thirty years and most of our remaining fleet is set to be taken offline in the next few years. Only Sizewell B, which opened in 1995, is planned to stay online past 2028. With the delay of the new Hinkley Point C reactor, when Sizewell B closes for maintenance in 2029, Britain will have no nuclear power whatsoever on the grid for the first time in more than 70 years.

This would be bad for energy security, bad for household bills, and bad for climate change, pushing up the amount of unabated gas we burn to compensate. It also puts the Government’s target for a clean energy grid by 2030 at risk, with the recent report from the National Energy System Operator finding that this target requires there being between 3.6 GW and 4.1 GW of nuclear power on the grid.

But there is a solution. At Britain Remade we’re calling on the Government to safely extend the life of our existing fleet of Advanced Gas-cooled Reactors (AGRs). Our analysis shows that extending the use of just two of our AGRs, Heysham 2 and Torness power stations, and finishing just 1 unit of Hinkley Point C would provide 5.3 GW of clean nuclear power, preventing the release of 8.8 million tonnes of carbon into the atmosphere, equivalent to taking 1.8 million petrol cars off the road.

That’s why we’re pushing the government to follow in the footsteps of countries like France and the United States and make the obvious, sensible and safe decision to extend the lifespan of our existing nuclear reactors. If you agree with us, back our campaign.

👩‍💻 Anastasia Bektimirova, Researcher

I spent a fascinating couple of days last week at the events of the inaugural AI for Science Forum, co-hosted by Google DeepMind and The Royal Society. True to the spirit of open science, the organisers recorded and uploaded the Forum’s sessions here. The themes that kept on coming up were an evolving notion of what it means to be a scientist, multidisciplinarity, and the changing ways of organising science. But unambiguously, the keyword was: data – both as a bottleneck and an opportunity – for AI-driven scientific progress. As DeepMind’s VP of Science Pushmeet Kohli noted during his panel, “we have a path for discovery but a scarcity of input.”

This point holds for all fields but one – the social sciences. As Fabian Theis, Director of the Computational Biology Institute at TUM said, “we have all these different developed disciplines…they have datasets…but then they still see this [AI] as a bit of a mystic thing”. What is needed is “activation to embrace AI”. Later, the Government Chief Scientific Adviser Dame Angela McLean said, “it’s incredibly attractive to use these powerful tools to address natural science questions, but we should also be thinking how we can use these tools to completely revolutionise social sciences”. I couldn’t agree more, with the caveat that AI is not the only available hammer of this revolution.

Social sciences are not uniform. Certain fields and subfields are inherently more receptive to methodological innovation than others. Updating university social science curriculum to better align it with methodological frontiers can only go so far. The “digital revolution” in the social sciences, prompting increased use of computational research methods, has been around for a while, with uneven spread. This suggests that beyond methodological innovation, deeper shifts need to happen too. They may need to be more cultural in nature, which makes them harder to achieve or incentivise.

With the exception of certain technically advanced camps in traditionally quantitatively strong fields, such as economics and political science, AI as a research tool remains a largely untapped opportunity in the social sciences. DeepMind’s recent blog on AI for science is a telling sign. It is a brilliant read, thoroughly exploring AI opportunities and challenges across a wide spectrum of fields, but overlooks the social sciences, even though DeepMind itself has produced impressive work in this space. Many areas discussed in the piece, such as problem selection and interdisciplinarity, certainly apply to social science research, yet it doesn’t spell out the words social science once. Personally, having attended several events in recent months on the topic, I’m left with no doubt that the future of the AI-powered social science revolution is bright.

⚖️ Philip Salter, Founder

The prosaic world British regulatory reform has caught the attention of the US under the stunning headline (which deserves to be written out in full): Shock UK Regulatory Coup Gives Government Sweeping Control Over US Tech: London's digital markets, competition and consumers act gives regulators the power to stop any acquisition, anywhere, for basically any reason they want.

Ashley Rindsberg is dissecting the Digital Markets, Competition and Consumers (DMCC) Act which is coming into force and gives the Competition and Markets Authority (CMA) significant new powers. Despite the headline, this isn’t clickbait – Rindsberg has a point. More precisely, he has lots of points, but I want to focus on the one that really matters to entrepreneurs that I talk with.

As ​​Rindsberg writes, “the DMCC gives the CMA broad new powers to intervene with mergers that target what it calls ‘killer acquisitions.’ While the language makes it sound as if only massively disruptive technologies fall into the category, the letter of the law reveals the opposite to be true. Instead, under the killer acquisition provision, the UK now has the ability to intervene in ‘no-increment’ mergers – i.e. deals that don’t increase the market share of either party, which can be manifestly non-competitive with each other.”

Back in 2021, we joined forces with the International Centre for Law and Economics to write a sober assessment paper on the risks of the DMU. Central to the argument in Conflicting Missions is that startups depend on acquisitions. While some don’t want to admit it, being bought is the main way entrepreneurs and venture capital investors are paid for their hard work and investment. The harder it is to sell your company, the harder it is to make a return. The paper cites empirical evidence that venture capital activity grows when countries enact pro-takeover laws, and declines when anti-takeover laws are introduced.

As our paper and Rindsberg’s article make clear – there is a huge amount of uncertainty about how this will play out. Entrepreneurs and investors are already burdened with enough of that.

---