The global AI race is no longer a contest of scale alone – it is becoming a competition of strategy. That sentiment was at the heart of the message delivered by Llion Jones, one of the original authors behind the 2017 Transformer paper and a key figure in the breakthrough that reshaped modern artificial intelligence. At YourNewsClub, we see his warning as more than a regional argument: it is a roadmap for countries that cannot, and should not, attempt to match the hyper-scaling giants of the United States and China.
Jones, now living in Tokyo and leading Sakana AI, argues that neither Wales nor the UK can compete in a race defined by trillion-dollar data centers, vast GPU clusters and institutions built explicitly for large-model industrialization. But instead of treating that reality as a defeat, he calls it an opportunity – an invitation to develop the kind of research that doesn’t rely on massive infrastructure but on intellectual differentiation. We often describe this as “strategic courage”: choosing the game you can win, not the one you are destined to lose.
YourNewsClub analyst Jessica Larn, whose specialization lies in macro-level technological policy, emphasizes that scale is not only a technical category, but an infrastructural and political one: “When elite decisions crystallize into computing architecture, competing with hyperscalers becomes an exercise in power, not innovation.” Her insight reframes Jones’s argument – nations must understand where they fit not in the tech market, but in the global computational hierarchy.
Owen Radner, our analyst focused on the physical and digital infrastructure of the AI era, draws a parallel with energy networks: “AI is the new power grid. Without cables, data nodes and electrical capacity, you don’t have a route into the future.” His framing makes the Welsh debate over grid upgrades especially urgent. Protests in central Wales against new transmission lines reflect local frustration, but they also endanger the region’s ability to attract data centers and research hubs – the very foundations of any credible AI ecosystem.
Meanwhile, the UK government has been attempting to accelerate AI adoption through business initiatives, regional AI growth zones and partnerships with major tech companies. Business Secretary Peter Kyle argues that even small increases in productivity – a mere 1% in the SME sector – could boost economic activity by over £240 billion. But as we at YourNewsClub note, projections collapse without infrastructure: Wales’s grid remains underpowered, under-expanded and unable to support the energy needs of large-scale computational operations.
Jones’s critique of modern chatbots adds another layer to the debate. He argues that current systems have become excessively flattering – optimizing for user approval rather than truth or correction. “People like being told they are right,” he notes. “But that means models will mislead them.” At YourNewsClub, we regard this as a fundamental design flaw: an AI that only mirrors human biases cannot help societies make better decisions. Jones’s call for models capable of disagreement – capable of saying “no” – marks a shift toward more mature, less human-pleasing systems.
Wales now stands between two divergent paths. The first is to chase the leaders and fail. The second – the one Jones champions – is to build an environment where universities, labs and independent research groups can experiment freely, unconstrained by the gravitational pull of Big Tech. We believe this second path is not only viable but necessary for regions seeking technological sovereignty.
Our conclusion at Your News Club is straightforward: nations that stop trying to compete on sheer computational volume and instead invest in differentiated research, resilient infrastructure and intellectual autonomy will emerge stronger in the next phase of the AI era. The UK may never out-scale the U.S. or China – but it can outperform them in ideas, if it chooses its direction with precision and courage.