Somewhere in the Treasury, there is a civil servant whose job is to produce Britain’s growth plans. To generate PDF after PDF featuring stylised maps of the country, haloed with neon light, suggestive of energy, vigour and dynamism. To that nameless graphic designer, Rachel Reeves’s promise to make Britain a “defence industrial superpower” will have resounded like a bugle call. The rest of us may be a tad more sceptical.
It’s not just that the government’s mission until only a few days ago was to make us a clean energy superpower instead (pinky-purple wireframe wind turbine, very tasteful). It’s that, increasingly, the only thing Britain leads the world at is saying we’ll lead the world. In fact, “world-leading” has been used in 2,386 prime ministerial announcements and press releases in recent years. Not least by Boris Johnson.
Of course, it’s a good and necessary thing that the government is spending more on defence. But that extra £25.3 billion, spread over five years, represents only 0.35 per cent of Britain’s projected £7.2 trillion outlay over that period. So, while it will certainly sustain the defence sector, it is unlikely to move the economic needle.
But there’s another problem. To be a “defence industrial superpower”, you don’t just need a defence industry — you need industry, full stop. China’s naval shipbuilding capability is estimated to be 232 times that of the United States largely because it dominates civilian shipbuilding too.
Without that industrial base, there is the risk that most of the parts for those fancy jets, tanks and drones, and even more of the raw materials, are coming from states that are not 100 per cent on your side. So how’s that going? Well, last week British Steel — already Chinese-owned — announced that it might have to close its remaining blast furnaces at Scunthorpe, which would make us the only G7 country that cannot produce virgin steel.
It’s not alone. Car production in the UK, traditionally a key strength and key export, has halved over the past six years, even before Trump’s tariffs. North Sea oil and gas is collapsing. Refineries are closing: Sir Jim Ratcliffe, whose firm Ineos is closing the oil refinery and ethanol plant at its Grangemouth site, says we are “witnessing the extinction” of the chemicals sector. The ceramics industry recently announced a “rescue plan” to address “unprecedented challenges”. Britain’s foundries, whose output has already fallen by 85 per cent since 2000, face further closures.
In this environment, press releases from industry groups drip with desperation. Make UK, the umbrella body for manufacturing, complains that “manufacturers feel like they are currently wading through treacle, facing barriers and increased costs being imposed on them at every turn”. The chemicals industry warns of energy bills that are 400 per cent higher than in America, “unsustainable costs” and a “hostile policy agenda”, asking — as Ratcliffe did — whether the country wants it to exist at all.
A large part of the problem is, as the above suggests, energy prices and charges — which are being made worse by Ed Miliband’s desperate dash to decarbonise the grid by 2030, irrespective of the costs. But as I’ve pointed out before, we’ve also designed our entire net-zero system to count emissions in the UK only. So if that British Steel plant shuts down, and we start importing from China or India, it counts as a carbon win even if their steel is produced using the dirtiest coal on the planet and with none of the green levies we apply here.
In recent decades the proportion of our carbon emissions incurred through imports has duly been rising. In the process, as Kemi Badenoch pointed out recently, we have become overwhelmingly dependent on China to reach net zero — and, increasingly, to keep the lights on — since it dominates the production of solar panels, wind turbines, electric cars and more besides.
But it’s not just about energy or climate. Britain has displayed a slapdash indifference to its manufacturing sector. We permitted the ridiculous ossification of the planning system, making it hard to build not only houses but factories, laboratories and the roads to connect them. George Osborne funded his corporate tax cuts by slashing the capital allowances that manufacturers relied on to invest.
Year after year, the country that invented the Industrial Revolution sets new records for how small its manufacturing sector can shrink as a share of GDP. As of 2023, it was 8.3 per cent — half of the level in 1990. Of course, manufacturing is struggling across the West: just look at Germany. China is eating everyone’s lunch. But our sector is a smaller part of the economy than in any of our main competitors.
We are also well below our peers in R&D investment. A few years ago the analyst Rian Chad Whitton pointed out that only 121 British companies appeared in a list of the world’s top 2,500 spenders on R&D. In the 2024 edition of the same index, that had fallen to 63 out of the top 2,000. And many were firms such as AstraZeneca, GlaxoSmithKline or Rio Tinto, which have a corporate presence in the UK but operate and innovate across the globe.
Britain, in short, has been conducting a vast experiment to see how much a country can grow without making or building stuff — creating an economy that privileges not only services over manufacturing, but the intangible over the actual.
In a fascinating piece the other day, Whitton showed that between 2000 and 2023, Britain’s reliance on imports for basic materials grew hugely, even as our consumption of those materials shrank. We use much less cement than our neighbours, much less steel, much less tarmac, much less fuel, to the point where we look more like a Latin American than a western European country in the data.
Whitton concludes that “the British economy’s ability to actually process new material, from gravel to advanced machinery, is internationally substandard”, and that we are “the most un-automated major country in the developed world”.
I wish ministers well with their plans to make us a defence superpower. But it’s hard to reconcile that ambition with our status as a country in which making things is often viewed as a second-class activity, which relies on other nations for a striking proportion of the physical goods it consumes, and which has adopted policies on energy, climate and planning that are actively making those problems worse.
America, under both Biden and Trump, has belatedly become serious about reindustrialising, in particular its military production. Far from being a “superpower”, Britain is scrabbling to catch up in a world in which all too many of our supply chains, whether for energy, technology or industry, ultimately depend on the kindness of strangers.
It’s thought to be unlikely that chancellor Rachel Reeves will axe the programme altogether
Crackdown on R&D cash
Will Labour revise state help designed to foster research and development?
Could research and development (R&D) tax credits be a target for Rachel Reeves in her first Budget? The annual cost of the credits, paid to businesses investing in innovation, has increased to around £7.5bn a year. And amid warnings that not all claims are bona fide, that pot of cash could make a tempting target.
More than 55,000 small businesses received R&Dtax credits last year, thelatest figures from HM Revenue & Customs show, underlining the value of this scheme to large numbers of companies. However, the scheme has already undergone substantial changes, with reforms introduced in April aimed at simplifying the system and cracking down on ineligible claims.
Recent reform
The new arrangements have merged the two separate schemes that discriminated between claims made by small and larger businesses. However, the principle remains the same: if your business invests in innovation, it should be able to offset some the cost of that investment against its corporation tax. And if you’re not paying corporation tax because your business is not currently profitable, you should still be eligible for support.
The rules get quite technical, but the relief is a valuable one, enabling you to claim back up to 27% of your innovation costs (depending on your circumstances) under the new merged scheme. Claims can be made in relation to innovation costs incurred in your past two accounting years.
Importantly, innovation has a broad meaning under the scheme. It might be that your business is investing in trying to make some sort of advance in science or technology. Or you may be seeking to overcome some sort of scientific or technological uncertainty. That brings a broad range of work into play. It’s not only major technology breakthroughs that count, but also investments in process or development – a company that finds a new way to run an operation or execute its production, say, may be eligible to claim. While HMRC’s figures show that companies in the information and communications sector account for more R&D tax credits than any other, it also receives plenty of claims from manufacturers, professional services companies and the construction industry.
The bottom line is that if your firm is pursuing innovation of any kind, it is worth looking in to whether you are eligible for support. And while it seems unlikely that the chancellor would axe the scheme altogether, she may seek to make further changes to the rules. It therefore makes sense to assess your position now, if only to understand how the budget affects you.
That said, tread carefully with claims. In recent years, the government has become increasingly concerned about the quality of claims – and fraud – and HMRC has been scrutinising filings more closely. The tax authority even has powers to claw back credits it decides should not have been paid, with a growing number of small businesses handed bills for thousands of pounds. Some of those demands relate to claims made several years ago.
It’s therefore imperative not to leave yourself vulnerable to a difficult inquiry from HMRC. It may be a good idea to take professional advice from an accountant or a tax-credits specialist before proceeding with a claim. But work with an adviser you trust. A small cottage industry has grown up around the tax-credits sector, with firms making bold claims about how much support they can secure for businesses. They will typically take a sizeable chunk of this cash – and if HMRC does subsequently investigate your case, it may be difficult to pursue the adviser.
Small firms will be feeling underwhelmed by the Autumn Statement
SOME SMALL COMPANIES MAY STRUGGLE TO COPE WITH THE RISE IN THE MINIMUM WAGE
S
mall businesses hoping for extra support from last week’s Autumn Statement may have been disappointed. While the Chancellor unveiled several eye-catching initiatives, small and medium-sized enterprises (SMEs) can be forgiven for feeling underwhelmed.
Take “full expensing”, the Chancellor’s centre-piece announcement. It makes permanent a tax break that enables SMEs to claw back 25p of corporation tax for every £1 they invest in areas such as IT, machinery and equipment, by deducting the cost of these items from the profits they declare to HM Revenue & Customs.
That sounds generous, but most small businesses won’t make use of full expensing. That’s because there is already an annual investment allowance that offers exactly the same benefits on investment of up to £1m a year – more than most SMEs are likely to invest. If you are using the AIA, however, remember that this tax break can only be claimed in the year in which the asset is purchased.
SIMPLIFIED R&D REGIME
Similarly, the Chancellor’s announcement that he is to merge the two schemes offering tax reliefs to businesses that spend money on research and development (R&D) will simplify this system.
But it is larger businesses that will benefit, since they will now also qualify for the 20% tax credits that only smaller enterprises previously claimed. Moreover, the merged schemeis more complex than the previous arrangement for smaller businesses, which will mean an extra administrative burden for some.
That said, SMEs that are currently loss-making will still qualify for more generous payouts from the R&D tax-credits scheme. And the Chancellor has announced a relaxation of the rules on this part of the scheme, which should help more businesses to claim.
Other announcements also look pretty modest. For example, the Chancellor reiterated his determination to crack down on late payments, which continue to cause many smaller businesses major problems with cash flow. In practice, however, the only changes here are that the state will require businesses bidding for large public-sector contracts to show they pay their own invoices within an average of 55 days. That will only help smaller firms at the margins.
Meanwhile, the announcement that the minimum wage will increase to £11.44 an hour for workers over 23, and be extended to include 21- and 22-year-olds for the first time, will raise costs for all employers. Smaller firms may find it harder to cope with the 9.8% increase, which is well ahead of both inflation and the current rate at which national average earnings are increasing. However, the Chancellor did at least offer more support for the self-employed. The abolition of Class 2 national insurance contributions will save the average self-employed worker £192 a year – and simplify their taxes. And a one-percentage point reduction in the Class 4 rate of national insurance will be even more valuable when it comes into effect in April 2024.
Some self-employed workers will also benefit from the fact that the government won’t now extend the Making Tax Digital system to those earning less than £30,000 a year. The scheme requires self-employed workers and small businesses to file tax returns through software and digital platforms approved by HMRC. But those earning more modest sums won’t now have to switch to digital returns unless they want to.
It is already making research faster, better, and more productive
image: shira inbar
Listen to this story.Enjoy more audio and podcasts on iOS or Android.
In 2019, scientists at the Massachusetts Institute of Technology (mit) did something unusual in modern medicine—they found a new antibiotic, halicin. In May this year another team found a second antibiotic, abaucin. What marked these two compounds out was not only their potential for use against two of the most dangerous known antibiotic-resistant bacteria, but also how they were identified.
In both cases, the researchers had used an artificial-intelligence (ai) model to search through millions of candidate compounds to identify those that would work best against each “superbug”. The model had been trained on the chemical structures of a few thousand known antibiotics and how well (or not) they had worked against the bugs in the lab. During this training the model had worked out links between chemical structures and success at damaging bacteria. Once the ai spat out its shortlist, the scientists tested them in the lab and identified their antibiotics. If discovering new drugs is like searching for a needle in a haystack, says Regina Barzilay, a computer scientist at mit who helped to find abaucin and halicin, ai acts like a metal detector. To get the candidate drugs from lab to clinic will take many years of medical trials. But there is no doubt that ai accelerated the initial trial-and-error part of the process. It changes what is possible, says Dr Barzilay. With ai, “the type of questions that we will be asking will be very different from what we’re asking today.”
Drug discovery is not alone in being jolted by the potential of ai. Researchers tackling many of the world’s most complicated and important problems—from forecasting weather to searching for new materials for batteries and solar panels and controlling nuclear-fusion reactions—are all turning to ai in order to augment or accelerate their progress.
The potential is enormous. “ai could usher in a new renaissance of discovery,” argues Demis Hassabis, co-founder of Google DeepMind, an ai lab based in London, “acting as a multiplier for human ingenuity.” He has compared ai to the telescope, an essential technology that will let scientists see farther and understand more than with the naked eye alone.
Where have you been?
image: the economist
Though it has been part of the scientific toolkit since the 1960s, for most of its life ai has been stuck within disciplines where scientists were already well-versed in computer code—particle physics, for example, or mathematics. By 2023, however, with the rise of deep learning, more than 99% of research fields were producing ai-related results, according to csiro, Australia’s science agency (see chart). “Democratisation is the thing that is causing this explosion,” says Mark Girolami, chief scientist at the Alan Turing Institute in London. What used to require a computer-science degree and lines of arcane programming languages can now be done with user-friendly ai tools, often made to work after a query to Chatgpt, Openai’s chatbot. Thus scientists have easy access to what is essentially a dogged, superhuman research assistant that will solve equations and tirelessly sift through enormous piles of data to look for any patterns or correlations within.
In materials science, for example, the problem is similar to that in drug discovery—there are an unfathomable number of possible compounds. When researchers at the University of Liverpool were looking for materials that would have the very specific properties required to build better batteries, they used an ai model known as an “autoencoder” to search through all 200,000 of the known, stable crystalline compounds in the Inorganic Crystal Structure Database, the world’s largest such repository. The ai had previously learned the most important physical and chemical properties required for the new battery material to achieve its goals and applied those conditions to the search. It successfully reduced the pool of candidates for scientists to test in the lab from thousands to just five, saving time and money.
The final candidate—a material combining lithium, tin, sulphur and chlorine—was novel, though it is too soon to tell whether or not it will work commercially. The ai method, however, is being used by researchers to discover other sorts of new materials.
What did you dream?
ai can also be used to predict. The shapes into which proteins twist themselves after they are made in a cell are vital to making them work. Scientists do not yet know how proteins fold. But in 2021, Google DeepMind developed AlphaFold, a model that had taught itself to predict the structure of a protein from its amino-acid sequence alone. Since it was released, AlphaFold has produced a database of more than 200m predicted protein structures, which has already been used by over 1.2m researchers. For example, Matthew Higgins, a biochemist at the University of Oxford, used AlphaFold to figure out the shape of a protein in mosquitoes that is important for the malaria parasite that the insects often carry. He was then able to combine the predictions from AlphaFold to work out which parts of the protein would be the easiest to target with a drug. Another team used AlphaFold to find—in just 30 days—the structure of a protein that influences how a type of liver cancer proliferates, thereby opening the door to designing a new targeted treatment.
AlphaFold has also contributed to the understanding of other bits of biology. The nucleus of a cell, for example, has gates to bring in material to produce proteins. A few years ago, scientists knew the gates existed, but knew little about their structure. Using AlphaFold, scientists predicted the structure and contributed to understanding about the internal mechanisms of the cell. “We don’t really completely understand how [the ai] came up with that structure,” says Pushmeet Kohli, one of AlphaFold’s inventors who now heads Google DeepMind’s “ai for Science” team. “But once it has made the structure, it is actually a foundation that now, the whole scientific community can build on top of.”
ai is also proving useful in speeding up complex computer simulations. Weather models, for example, are based on mathematical equations that describe the state of Earth’s atmosphere at any given time. The supercomputers that forecast weather, however, are expensive, consume a lot of power and take a lot of time to carry out their calculations. And models must be run again and again to keep up with the constant inflow of data from weather stations around the world.
Climate scientists, and private companies, are therefore beginning to deploy machine learning to speed things up. Pangu-Weather, an ai built by Huawei, a Chinese company, can make predictions about weather a week in advance thousands of times faster and cheaper than the current standard, without any meaningful dip in accuracy. FourCastNet, a model built by Nvidia, an American chipmaker, can generate such forecasts in less than two seconds, and is the first ai model to accurately predict rain at a high spatial resolution, which is important information for predicting natural disasters such as flash floods. Both these ai models are trained to predict the weather by learning from observational data, or the outputs of supercomputer simulations. And they are just the start—Nvidia has already announced plans to build a digital twin of Earth, called “Earth-2”, a computer model that the company hopes will be able to predict climate change at a more regional level, several decades in advance.
Physicists trying to harness the power of nuclear fusion, meanwhile, have been using ai to control complex bits of kit. One approach to fusion research involves creating a plasma (a superheated, electrically charged gas) of hydrogen inside a doughnut-shaped vessel called a tokamak. When hot enough, around 100m°C, particles in the plasma start to fuse and release energy. But if the plasma touches the walls of the tokamak, it will cool down and stop working, so physicists contain the gas within a magnetic cage. Finding the right configuration of magnetic fields is fiendishly difficult (“a bit like trying to hold a lump of jelly with knitting wool”, according to one physicist) and controlling it manually requires devising mathematical equations to predict what the plasma will do and then making thousands of small adjustments every second to around ten different magnetic coils. By contrast, an ai control system built by scientists at Google DeepMind and epfl in Lausanne, Switzerland, allowed scientists to try out different shapes for the plasma in a computer simulation—and the ai then worked out how best to get there.
Automating and speeding up physical experiments and laboratory work is another area of interest. “Self-driving laboratories” can plan an experiment, execute it using a robotic arm, and then analyse the results. Automation can make discovering new compounds, or finding better ways of making old compounds, up to a thousand times faster.
You’ve been in the pipeline
Generative ai, which exploded into public consciousness with the arrival of Chatgpt in 2022 but which scientists have been playing with for much longer, has two main scientific uses. First, it can be used to generate data. “Super-resolution” ai models can enhance cheap, low-resolution electron-microscope images into high-resolution ones that would otherwise have been too expensive to record. The ai compares a small area of a material or a biological sample in high resolution with the same thing recorded at a lower resolution. The model learns the difference between the two resolutions and can then translate between them.
And just as a large language model (llm) can generate fluent sentences by predicting the next best word in a sequence, generative molecular models are able to build molecules, atom by atom, bond by bond. llms use a mix of self-taught statistics and trillions of words of training text culled from the internet to write in ways that plausibly mimic a human. Trained on vast databases of known drugs and their properties, models for “de novo molecular design” can figure out which molecular structures are most likely to do which things, and they build accordingly. Verseon, a pharmaceutical company based in California, has created drug candidates in this way, several of which are now being tested on animals, and one—a precision anticoagulant—that is in the first phase of clinical trials. Like the new antibiotics and battery materials identified by ai, chemicals designed by algorithms will also need to undergo the usual trials in the real world before their effectiveness can be assessed.
A more futuristic use for llms comes from Igor Grossmann, a psychologist at the University of Waterloo. If an llm could be prompted with real (or fabricated) back stories so as to mirror accurately what human participants might say, they could theoretically replace focus groups, or be used as agents in economics research. llms could be trained with various different personas, and their behaviour could then be used to simulate experiments, whose results, if interesting, could later be confirmed with human subjects.
llms are already making scientists themselves more efficient. According to GitHub, using tools like its “Copilot” can help coders write software 55% faster. For all scientists, reading the background research in a field before embarking on a project can be a daunting task—the sheer scale of the modern scientific literature is too vast for a person to manage. Elicit, a free online ai tool created by Ought, an American non-profit research lab, can help by using an llm to comb through the mountains of research literature and summarise the important ones much faster than any human could. It is already used by students and younger scientists, many of whom find it useful to find papers to cite or to define a research direction in the face of a mountain of text. llms can even help to extract structured information—such as every experiment done using a specific drug—from millions of documents.
Widening access to knowledge within disciplines could also be achieved with ai. Each detector at the Large Hadron Collider at cern in Geneva requires its own specialised teams of operators and analysts. Combining and comparing data from them is impossible without physicists from each detector coming together to share their expertise. This is not always feasible for theoretical physicists who want to quickly test new ideas. Miguel Arratia, a physicist at the University of California, Riverside, has therefore proposed using ai to integrate measurements from multiple fundamental physics experiments (and even cosmological observations) so that theoretical physicists can quickly explore, combine and re-use the data in their own work.
ai models have demonstrated that they can process data, and automate calculations and some lab work (see table). But Dr Girolami warns that whereas ai might be useful to help scientists fill in gaps in knowledge, the models still struggle to push beyond the edges of what is already known. These systems are good at interpolation—connecting the dots—but less so at extrapolation, imagining where the next dot might go.
And there are some hard problems that even the most successful of today’s ai systems cannot yet handle. AlphaFold, for example, does not get all proteins right all the time. Jane Dyson, a structural biologist at the Scripps Research Institute in La Jolla, California, says that for “disordered” proteins, which are particularly relevant to her research, the ai’s predictions are mostly garbage. “It’s not a revolution that puts all of our scientists out of business.” And AlphaFold does not yet explain why proteins fold in the ways they do. Though perhaps the ai “has a theory we just have not been able to grasp yet,” says Dr Kohli.
image: shira inbar
Despite those limitations, structural biologists still reckon that AlphaFold has made their work more efficient. The database filled with AlphaFold’s protein predictions allows scientists to work out the likely structure of a protein in a few seconds, as opposed to the years and tens of thousands of dollars it might have taken otherwise.
And speeding up the pace of scientific research and discovery, making efficiencies wherever possible, holds plenty of promise. In a recent report on ai in science the oecd, a club of rich countries, said that “while ai is penetrating all domains and stages of science, its full potential is far from realised.” The prize, it concluded, could be enormous: “Accelerating the productivity of research could be the most economically and socially valuable of all the uses of artificial intelligence.”
Welcome to the machine
If ai tools manage to boost the productivity of research, the world would no doubt get the “multiplier for human ingenuity” predicted by Dr Hassabis. But ai holds more potential still: just like telescopes and microscopes let scientists see more of the world, the probabilistic, data-driven models used in ai will increasingly allow scientists to better model and understand complex systems. Fields like climate science and structural biology are already at the point where scientists know that complicated processes are happening, but researchers so far have mainly tried to understand those subjects using top-down rules, equations and simulations. ai can help scientists approach problems from the bottom up instead—measure lots of data first, and use algorithms to come up with the rules, patterns, equations and scientific understanding later.
If the past few years have seen scientists dip their toes into the shallow waters of ai, the next decade and beyond will be when they have to dive into its depths and swim towards the horizon.
