Rishi Sunak’s AI vision for the UK – doomed for failure?
By Dr Nasira Bradley
If Rishi Sunak is to be believed, Britain will be a “science and technology superpower” by 2030! Rishi Sunak thinks that Britain will harness AI — just like that! – and spur productivity and economic growth.
Labour likewise professes that “UK is well positioned to benefit from transformative AI and to lead the world to ensure AI is aligned with public interests”.
These statements beg the question ‘how’ is the UK going to become a science and technology super-power all of a sudden? One need only look at the leading technology global firms over the last few decades – to realise not a single UK leading global firm is to be found in the areas of social media platforms or database or cloud computing! Indeed, the UK has not produced the Facebooks, the Googles, the Amazons or the Apples! None, barr ARM over the last 40 years. Surely our politicians and policy makers understand that an ambition to become a leading technology superpower – requires some accompanying leading global technology firms. Very few would dispute this fact.
Yet, what is baffling, time and time again, is that British policy makers seem to miss this key ingredient … To dominate as an economic super-power, we don’t just need brilliant universities or nimble innovative start-ups , which are essential but not sufficient— we need to grow some of them into innovative giants!
America realised this early on shortly after Second World War. What has allowed the US to be a dominant superpower over last 50 years — are the inventions of ‘the computer, the microchip & the internet’ (Isaac Walterson)! But it’s not only the invention of these key technologies, it is the US giant firms that commercialised gains from these technologies that ensured America its super-power status!
More to the point, these inventions didn’t just happen! They were brought about through a powerful triangular alliance between government, academia and industry — an alliance carefully forged, mostly by Vannever Bush, Dean of MIT School of Engineering, founder of electronics company Raytheon and Americas top military science administrator during World War 2. Bush’s seminal paper “Science, the Endless Frontier” in 1945, laid the foundation for this triangular alliance persuading the government that through this path lay economic superiority — and thus military superiority for America’s edge in the world.
Fred Terman, oft referred to as ‘father of Silicon Valley’ was Dean of Engineering at Stanford post second world-war and influenced by Bush’s thinking in MIT, brought this triangular alliance thinking to Santa Clara County, later dubbed Silicon Valley. So, how were Silicon firms grown in the first instance? In fact, pentagon contracting was fundamental to the building up of the silicon industry in these post war years — a focus that Fred Terman, himself keenly encouraged, proactively laying this foundation between academia, industry and defence.
This role of military contracting is not readily acknowledged by entrepreneurs of the region — let alone well known. This is not to undermine the role of entrepreneurs, venture capital etc who also were and remain critical. Yet, the role of Pentagon and military contracting is not minor. Often downplayed by both industry itself and US policy makers — it was critical to the making of Silicon Valley in the 50’s and 60’s and even its remaking in the late 90’s. Even now it continues its silent but powerful presence… and why Graphcore matters!
Unless and until our policy makers truly start understanding how US built their own giant tech and emulate them to build our own technological giants — we don’t stand a chance of being any kind of AI superpower! So, what should UK do?
We only have to look at the top 20 or even top 50 innovative firms in the world, listed on the Innovation Scorecard for 2023 and see it is dominated by US firms — with Astrazenca in the top 20 & GSK in the top 50, as the two UK exceptions! In contrast, if we look at the list — Alphabet, Meta, Apple, Microsoft are all American firms created in the last 40 years, topping the list!
So, how then did US build its technology — and in particular what is its ‘secret sauce’ that built the first giant firms up? It is this ‘first’ build-up of new giant technological companies in new sectors that is the hardest, as thereafter it is much easier to self-generate the next and the next. What then allowed US to develop those first ‘giant firms’ post world war, predecessors to the current successful tech giants — be it in computers, or the microchip or the internet?
Apart from role of venture capitalists and entrepreneurs, Pentagon spending and in particular, defence contracting played a pivotal role in the build-up of the many industries crucial to the making of Silicon Valley in the 50’s and 60’s — microelectronics, missile, satellites, space, semiconductors and internet to name but a few.
It’s not that every firm that military contracting supported became a commerical success — it’s that they encouraged sufficient number of firms in a wide range of technologically key industries to grow, hoping to allow some American technology firms to make it as leaders on the global stage. Some of the firms did not make it beyond the military contracting world and broke up, yet often spawning new start-ups of which some did make it. Others have lasted to this day!
Let’s take an example from the semiconductor industry . The Semiconduuctor story of William Shokley, co-inventor of the transistor in Bell-labs is well known. Shokley, initially founded Shockley Semiconductors in 1955 in Palo Alto, his hometown. However, three years later 8 employees left him to start Fairchild Semiconductor, growing to compete successfully in the markets alongside Motorola and Texas Instruments. Fairchild and many of the other semiconductor companies produced transistors for military programmes in the 50’s, so much so that they nearly were responsible for half the revenue of the entire industry. This guarantee of revenue in its initial years as it built itself up — cannot be underestimated. It gives the company a chance to hire the people it needs, to develop systems and processes & to grow!
Indeed, Fairchild only entered the commercial markets in the 1960’s, where its stringent detail for quality honed in military contracting helped it compete successfully against established firms such as Texas Instruments. Significantly even after entering commercial markets, Fairchild remained an important military contractor, which allowed it to develop technologies yet retain proprietary rights of usage. What was even more important, that Fairchild employees like Robert Noyce and Gordon Moore and some other left to form Intel, Jerry Sanders left to form AMD and Charles Sporck left to head National Semiconductor. These insights are important — such knock-on effects of The Pentagon and military contracting were repeated across critical industries in Silicon Valley today.
Thus, if the UK truly wants to be an AI super-power, it has to forge this triangular alliance — and make sure the government helps build as many as possible AI firms, so that one of them may turn out to be a Fairchild, which later become the Intels or AMDs or Nvidias of AI.
Given this thinking — UK’s Exascale Supercomputer would have seemed an ideal opportunity for Rishi Sunak to put his words into action. Why not use the £900 million ($1.1 billion) Exascale supercomputer project in Bristol to support AI domestic firms? What is baffling, the project will support American firms — and leave out the only domestic AI firm, Graphcore!
In fact the UK’s announcement of working on a supercomputer comes a decade later than US! DARPA was already making bets in 2010 that GPU’s (graphic processing units) could rival the then-dominant CPU’s (central processing unit) and allocated 25 million to Nvidia to work on developing the technology for an exascale supercomputer, alongside other companies. The point was not about picking winners, rather making multiple bets on various potential technologies that could shape the future. In 2010, it was the question of CPUs versus possible GPUs, yet who knows what the future will bring — could be GPU’s versus IPU’s (intelligence processing units)? No-one can predict the future, the aim for the government should be to support multiple options of technologies and companies.
Thus, it makes it even more perplexing, the UK government is betting on technology. Since when do governments know which technology will succeed, that GPUs are going to be the processor in future for AI, not leaving any space for IPU’s — Graphcore’s technology! The government has awarded initial contracts only to University of Bristol and Hewlett Packard Enterprise (HPE), which will be building the supercomputer based on Nvida GPU’s. Although, it is true that today’s technology for training AI systems uses GPUs, it could be that the chips required for actual AI-solving of queries could turn out to be IPU, who knows … Definitely, not yet the industry, but most definitely, not the government!
The wisdom would have been to allow scope for both – more so, critical to use the opportunity to develop multiple domestic firms, including Graphcore.
Indeed, who knows whether Graphcore will make it or not, every firm that stakes its bet on innovation breakthrough runs a risk — Nvdia made a bet on CUDA a software toolkit in 2006, which Wall Street refused to recognize for 10 years, only to have it recognized now! Yet Nvidia has also made its share of wrong bets, exemplified in its failure to win in the smartphone market with its TEGRA processors. The point is that innovative firms will get some calls right and some wrong. Graphcore having been lauded a year ago, has suddenly found itself out of favour after its deal with Microsoft fell through — even if it is acknowledged that Microsoft’s deal with ChatGPT most probably shaped its move towards Nvidia.
More to the point, is that Graphcore (and any other domestic UK chip design firms, we should include them all, including startups!) could be the UK’s Fairchild, — to later lead to the Intel-equivalent next AI firm!
So, if Rishi Sunak is serious about UK’s vision for AI in the future — let’s take a leaf out of the US playbook and how it built up its own first tier of technological firms. Just like the Pentagon actively has used military contracting to support the early gestation of many technical startups, the UK can do the same — invest and encourage many startups, including Graphcore and one of them may just turn out to be the Fairchild that becomes the intel or Nvidias of tomorrow!
Thus conclusively, what I have tried to point out is that the US is not laissez-faire when it comes to technology or innovative firms! So, why is the UK shy in this regard? There is a very active industrial policy in the US, it’s just hidden in the Department of Defence (DoD), with its roots shaped by Vannever Bush and Fred Terman’s thinking.
Let’s not forget Terman’s words in 1946 “Industrial activity that depends on imported brains and second-hand ideas cannot hope to be more than a vassal that pays tribute to its overlords and is permanently condemned to an inferior competitive position” — let’s get our own first-hand firms!
Let the UK have its own ‘first-hand’ ideas and companies!
