The Alan Turing Institute has failed to develop modern AI in the UK

The UK’s flagship institute for artificial intelligence, the Alan Turing Institute, has been at best irrelevant to the development of modern AI in the UK. Along with the AI council, which advises the government on AI, the Turing has been completely blindsided by recent breakthroughs in artificial intelligence based on large language models (LLMs).

The institute’s Annual reports for the last four years do not refer to LLMs at all. There is no record of its website or Director mentioning them until a few months ago. It’s as if the most important development in the history of AI has completely passed it by.

What have they concentrated on instead? Their most popular blog post in 2022 was “Non-fungible tokens: can we predict the price they’ll sell for?”. Their top piece of content was “Data as an instrument of coloniality: A panel discussion on digital and data colonialism”. Do any AI specialists think this work is going to push the bleeding edge of AI research in the UK?

I’m not speaking with the benefit of hindsight: the Royal Statistical Society (RSS) submission to the AI Council Roadmap (which I prepared) argued that the most pressing issue was the resources required to develop large language models. This was in 2020, almost three years ago.

The AI Council Roadmap and the subsequent National AI Strategy failed to acknowledge the role of large language models, concentrating instead on technology such as digital twins. Digital twins technology divides opinion amongst the AI practitioners I speak to: either they think it’s completely irrelevant to AI or they have no opinion because they’ve never heard of it.

LLM-based AI models may have an effect on jobs as large as the industrial revolution. There are fears of a transfer of power from the individual to technology companies like OpenAI and Google, who develop closed AI models. The only way to avoid these negative effects—and participate in the economic growth that AI will engender—is to develop open source AI models which are openly available to the public.

A document from Google was leaked last week suggesting that open source AI projects are  fast becoming a threat to large tech companies. The gap between open source models and closed models ‘is closing astonishingly quickly’. It’s concerning that none of the open source LLM projects mentioned in this document arose in the UK.

There are thousands of experienced AI developers in the UK who could contribute to open source projects. They need funding and leadership. The French government have already realised this: in 2021 they supported the creation of the open source LLM, BLOOM. Other projects are arising elsewhere, such as the Large European AI Models (LEAM) initiative. While we face headwinds, there is no reason we couldn’t do the same in the UK.

Hardware is one hurdle, the United Kingdom has only two top-500 supercomputers suitable for large language model training, while France already has six. This is easily fixed with funding.

But the biggest hindrance is that open source is a complete blindspot for the UK AI establishment. Open source was not mentioned at all in the AI Council Roadmap and barely mentioned in the UK National AI Strategy. 

We argued that open source should be a key component of the UK AI Strategy in our September 2021 article ‘The UK AI Strategy: are we listening to the experts?’ The RSS sent a letter containing a more formal request to several ministers in 2022. There was no policy response. More accurately, there was no response of any kind.

The government launched an AI Task Force in March 2023, for which £100M has been earmarked. Again, open-source is not part of the strategy. From the press release, it’s clear this initiative is treading the same tired path, based on the Turing and consultancies who specialise in government work. This is unlikely to result in value to the public.

We should have been preparing the ground: we have lost three years of network building, skills development and hardware deployment. But an undignified scrabble is better than being left behind. The following needs to happen in the next weeks and months, not years:

1. The AI Task Force should be dedicated to open source initiatives to build LLMs and related datasets. This should include technology companies as core participants and should not be run by the Alan Turing Institute.

2. Government organisations need to urgently engage with the technical community in the UK, to start giving it a voice at the heart of government AI initiatives. This will help avoid future missteps.

3. A panel of technical AI specialists should be set up to advise the government on AI. This should *not* include VC investors, thought-leaders, philosophers etc. unless they have technical skills and experience.

Radical change is needed. It’s said that to be at the cutting edge you need to live in the future. It’s not clear the UK AI establishment is even living in the present.

___

*[edit: An earlier version of this article suggested that no major open source AI projects originated in the UK. While Stable Diffusion was officially open sourced by the Ludwig Maximilian University of Munich, it was supported by Stability AI, a company which is headquartered in the UK. We have changed the text to make it clear refers to LLM models only.]

---

Martin Goodson was the Chair of the RSS Data Science and AI Section (2019-2022). He is the organiser of the London Machine Learning Meetup, the largest network of AI practitioners in Europe, with over 11K members. He is also the CEO of AI startup, Evolution AI.

The views expressed are his own and do not necessarily represent those of the RSS