AI pioneer Lion Jones calls on UK to ‘be brave’ in technology race

A computer scientist who has revolutionized the use of artificial intelligence (AI) has called for the UK to be “brave” in pursuing the technology.

Lyon-Jones, from North Wales, who was part of the pioneering Google team that unlocked AI’s ability to react in a human-like manner, warned against trying to compete with “hyperscalers” in the US and China.

He said the “sycophantic” response of the current wave of AI chatbots needs to be corrected.

“We need to create AI that can actually disagree with you and actually correct you,” he said.

Jones is one of the co-authors of the 2017 paper “Attending Is All You Need,” which introduced the Transformer architecture with a “T” for ChatGPT. This was a breakthrough advancement that powers the generative AI systems that are currently transforming business and daily life.

Still based in Tokyo, he is currently the chief technology officer of Sakana AI, a company he founded with another former Google researcher.

His company focuses on speculative research into future uses of AI, which he claimed is “unique” as other big tech companies focus on monetizing the current wave of large-scale language models.

“I really think we’re unique in that respect at the moment,” he said in an exclusive interview with BBC Wales.

“I’m trying to create an environment where people are free to do the research they want, and I’m going to try to protect that as much as I can.”

Mr Jones said governments in Cardiff and Westminster needed to accept the reality of global competition that the US and China were winning and instead focus on developing different types of AI.

“I don’t think countries like Wales, or even the UK in general, can beat the hyperscalers, right? They can’t beat the US or China in scaling up AI.

“There’s no point in participating in that race because you can’t win.

“So what can you do? You can embrace the idea that you can do something different. It just requires you to be a little bit braver.”

He said: “What Wales can do is do what I’m trying to do: create an environment in Wales where businesses and academic institutions feel they have the resources and freedom to undertake more differentiated research.”

“It’s a bit of a long gamble, but if it pays off, we can get back to racing, right?”

While consumers can access AI services on their smartphones, businesses in Wales are already racing to integrate time-saving and productivity-enhancing technology into their operations.

Google hosted an AI event with the UK government in Cardiff to encourage businesses to understand the benefits of adopting new technology.

“AI is not one thing and can be used in so many different situations,” Business Secretary Peter Kyle MP said.

He added that some of the companies participating in the event “work on movie sets, some work in living environments, and some are working on digital transformation.”

“Once this is integrated into the overall economy, it will have a transformative effect on the economy. It means more routine tasks will be taken away and a lot of business operations will actually be made faster,” Kyle said.

He added that a 1% increase in the productivity of small and medium-sized enterprises would generate economic activity worth £240 billion.

AI growth zones have been announced for north and south Wales, and the Welsh Government’s Wales AI Plan promises to ensure artificial intelligence is used to “drive economic growth, strengthen public services… and equip people across Wales with the skills to thrive”.

This includes a commitment to invest in AI innovation and skills in the public sector and develop ethical guidance for using the technology.

Lyon-Jones, who grew up in Bangor and Abergynolwyn, Gwynedd, studied at the University of Birmingham and worked in the UK technology industry before being hired by Google.

In California, he was one of the pioneering team of researchers that put the “T” on ChatGPT.

T is the Transformer architecture that enables AI to learn patterns from vast amounts of data, unlocking the ability of generative AI to describe, reason, and create at the same level as humans.

He joined Google, moved to Japan, and eventually left the company to start his own deep learning AI business.

Jones said today’s AI assistants need to be less flattering to users, as he sees the world grappling with the technology he helped develop.

Scaling up AI research and development in Wales will require investment from universities and government.

However, the computing power required to process and host AI services also requires an electricity grid, which is not currently available in all areas of Wales.

There has been considerable protest in mid-Wales against plans to increase capacity on the electricity grid, with the operator’s intention to build towers across the region.

“Mid Wales is chronically underserved by modern electricity grid infrastructure,” said Stuart George, chief executive of Green GEN Cymru, which is trying to build the towers.

Despite fierce opposition in some areas, Mr George argued that the deployment of a more reliable electricity grid was essential to supplying existing and future demand.

“What we need to do better is tell the story of the benefits that come from this,” he said.

“There are even more benefits to be gained from modernizing grid infrastructure.

“Given what the future holds, such as electric cars and electric home heating systems, it is currently not possible for businesses to decarbonise and grow and develop within Mid Wales.”

Mr George warned Wales could “lose out” in the AI ​​race if it did not have the infrastructure to attract data centers and research facilities, and said politicians needed to make difficult decisions.

Efforts to decarbonize the electricity supply mean that “the necessary grid infrastructure improvements to enable the use of renewable energy are inevitable, which means big decisions have to be made and they have to be made quickly.”

While Rion-Jones’ research focuses on the future of AI, he had recommendations to address the current wave of AI chatbots.

“They’ve become very sycophantic,” he says.

AI models are now trained to give answers that satisfy the humans they are interacting with.

“Unfortunately, humans love to be told they are right.

“This means that many people can be misled by the fact that you can say: [the AI chatbot] Almost anything, they just say, “Yes, you’re right, that’s great, that’s a really great idea.”

“So we need to solve that problem. We need to create an AI that can actually disagree with you and actually correct you.”