BritGPT: why SMEs need to be included in the UK’s plan to become an AI hub

The UK government is set to invest £900m in an exascale computer as part of its national artificial intelligence strategy.

Capable of carrying out more than one billion billion simple calculations a second, it would be used to train complex AI models so the UK can build its very own BritGPT.

The urgency of constructing a homegrown GPT is characteristic of an intensifying technological ‘arms’ race, where countries are competing to lead the AI revolution.

However, as the UK gets caught up in the politics of regulation and innovation, SMEs are increasingly being cast aside.

The dependence problem

BritGPt is, in principle, designed to maximise the country’s potential in AI and to give tools to researchers to better understand new drugs and climate change. However, there is also an existing political undercurrent.

According to Haydn Belfield, author of ‘ ‘Great British Cloud and BritGPT’ ’, our progress is dependent on a vulnerable and expensive AI supply chain that encompasses chip designers, data centres and AI developers in Silicon Valley.

Foundation models are expensive to train. Based on Belfield’s research, anything from £10 to £100 million just in computer costs with the potential to increase to the £1 to £10 billion range.

As a result, there’s been a near monopolisation of the provision of foundation models. The ‘Big Three’ AI cloud providers are all based in the US – Amazon Web services, Microsoft Azure, and Google Cloud.

Whilst the US is a strategic ally, logistically, problems could still arise. For instance, undersea cables could be damaged, prices could hike, or there simply could be a lack of oversight and control.

Although building cloud capacity that is comparable with the Big Three is not an overnight task, the UK can do two things to keep itself competitive.

Firstly, it can build up publicly owned cloud capacity that breaks a potential dependency cycle with the US. Secondly, and most importantly for SMEs, it can play to its key strengths in chip design, foundational model training and fine-tuning.

The role of SMEs in the UK’s AI Revolution

The AI business population in the UK is overwhelmingly comprised of startups or small businesses. Of the 3,710 UK companies that are currently registered in the country, 28% were small businesses and 60% were micro businesses.

The problem? 71% of all AI revenue (£7.6bn) was generated by large firms despite just making up 4% of the AI business population. On the other hand, SMEs companies together account for just over a quarter (£2.8bn) of AI revenue.

The skewed contribution between SMEs and big tech is indicative of how disparately capital is being allocated. Worryingly, it shows the risk aversion that continues to handicap SMEs seeking venture capital investment in the UK.

As a result, AI SMEs are being forced to look across the Atlantic for funding. However, rather than a recent problem, this is becoming an ongoing trend for startups in the country.

Autonomy, DeepMind, SwiftKey, and VocalIQ are all British AI and machine learning startups bought by US tech giants like Google and Microsoft.

The technologies these SMEs developed displayed potential to be market leaders. However, to grow, they required high levels of investment that the UK market was simply not offering. In the case of DeepMind, whilst Google acquired the company, a a team remained in London.

In other words, the UK has no shortage of AI talent. Some of the best universities of the world are based in Britain and it has one of the most vibrant environments for startups.

However, to avoid this potential brain drain, the UK needs to play to its strategic advantages in AI and give the right support to its startups that are helping drive the artificial intelligence revolution.

Whilst BritGPT is a step in the right direction to be technologically competitive in AI against giants like China or the US, the rollout needs to be more nuanced.

Otherwise, the UK risks falling victim to the growing monopolisation of funding and resources for big tech companies whilst startups struggle to find finance on home turf.