Unpacking the UK’s Newly Announced Centre on Artificial Intelligence

Few details about the planned UK defence centre on artificial intelligence (AI) have emerged since 19 November when the prime minister announced its intended formation. Nor is it clear whether it was the Cabinet Office and Downing Street or the Ministry of Defence itself that was the driving force behind the proposal, and it is not known whether the centre will reside within the defence organisational structure or be co-located with another department. As a result, all we can do at this stage is offer some suggestions about the functions the centre could perform and raise questions about its organisation and structure.

What Is It?

A few introductory lines about AI are needed. At its heart, it involves the use of computers for processing information to improve decision-making (namely suggesting choices that have a better chance of success and to do so more rapidly).

There are four elements in AI development. The first is categorised data, pieces of information that are sufficiently similar in nature that they can be meaningfully aggregated: every murder is different, but a murder is a murder. The second is computer hardware to store and process data, an area where technology is moving very quickly. The third comprises the algorithms – the software instructions telling the computer how to deal with the data. Finally, the fourth consists of the human subject matter experts responsible for specifying which data categories may be relevant to a particular situation: we cannot describe every aspect of anything, so selection is always needed. 

More advanced forms of AI are capable of ‘learning’ from additional data inputs and the feedback received about the outcomes of their responses. Rapid advances in computers’ capacity to store and process data, and in programmers’ skills in writing algorithms to guide computers in the processing of the data, are constantly transforming capabilities in this sector.

The Centre’s Potential Responsibilities

Significantly, AI is a particular field of knowledge that can be linked to a multitude of individual problems. Thus, a central challenge for the new body should be the extent to which the enhanced adoption of AI should be largely a bottom-up process in which multiple initiatives about specific issues come from below, or a top-down activity where centralised direction comes from higher up the decision chain. In the UK, as in the US, initiative on how defence money should be used lies with the individual services rather than any pan-defence body. And there is no assurance that the budget holders will understand or respond wisely to what AI could offer.

Five useful lines of duty can be discerned for an AI centre focused on defence in the UK, most of which resonate with defined areas of concern of the American Joint Artificial Intelligence Center (JAIC) set up in 2018.

The UK centre needs to be an educational and promotional body working to attract those most capable of applying themselves to AI in the defence context. However, it must also enhance basic understanding of AI’s potential and limits throughout the entire defence ecosystem. AI offers the potential for useful decision support capabilities; potential that will not be taken up unless relevant beneficiaries are aware of its possibilities. The US JAIC argues that building such expertise in many defence bodies will involve ‘culture change’ and envisages sending out AI evangelists. Education and promotion must include emphasis that most AI diagnoses of real-world situations are probabilistic and so may prove inaccurate: AI needs intelligent users.

Second, the centre could build understanding of the importance of data as the foundation for AI applications. Of the four AI elements noted above, arguably the most problematic at this stage is the first. Thus, the JAIC stresses that ‘data is at the heart of every AI project. We are addressing a wide range of data challenges, to include data collection, access, quality, common formats, ownership, and protections’.

The chief technical officer of the JAIC, Nand Mulchandani, asserts that AI is the new oil and that whichever government or country gets the best datasets will emerge as the leader in AI. Similarly, the UK's new centre should promote the generation of curated and interoperable data which, when appropriate, can be assembled in different packages to address more than one problem. Every problem should not have to collect its own dataset. Information packages defined for one issue (such as the potential reduction of injuries on training exercises) could well be valuable for another matter (for instance, the combinations of factors that come together to cause a trained soldier to leave the army). 

The JAIC has committed to the generation of standards and tools for data and AI that can be used to facilitate AI advances at lower levels. It is working on a Joint Common Foundation (JCF), a set of cloud-based tools to help organisations across the Department of Defense organise their data and develop their own applications. Its UK equivalent could play a similar role and there is clearly scope for collaboration in this space, not least because the US is interested in getting access to a large data pool, including material held by allies.

The UK centre could also take responsibility for the physical infrastructure, including the misleadingly named ‘cloud’, for the storage and processing of data. In contractual terms, this has proved very problematic in the US as protests about the award of the JEDI (Joint Enterprise Defense Initiative) project have stalled progress. To ensure the security of its data and avoid similar problems, the hardware and infrastructure for defence data should be located in the UK.

Finally, the forthcoming centre in the UK should lead on the country’s debates concerning the ethics, regulation and management of AI in defence; a much wider set of issues than whether autonomous destructive and lethal systems should be developed or who or what is responsible for choices made on the basis of AI process. There are also procurement challenges to be addressed and more thought is needed on the relative responsibilities of government and the private sector.

The US centre also engages in research to generate solutions to defined problems, but in this respect the UK government may doubt its capacity to attract the right talent, so the government may prefer to outsource research requirements.

In all its UK activities, the planned centre should develop a positive and complementary relationship with existing government defence bodies which are already wrestling with AI as part of their agendas, including the Development, Concepts and Doctrine Centre, the AI Centre of Expertise under Defence Digital, Dstl’s AI Laboratory, the Strategic Command’s jHub, the Army’s 6 Division, the RAF’s Rapid Capabilities Office and the Navy’s Digital Services piece.

The new centre should be staffed by AI experts operating as a team with those familiar in the broadest sense with how defence works and military problems. It will also need to develop relationships with public and private AI centres in the UK. In terms of command, the US centre is headed by a military officer with a civilian chief technical officer and it sits within the domain of the Department of Defense’s civilian chief information officer (CIO). The UK’s CIO is Charlie Forte, with his domain based at Corsham, as part of Strategic Command rather than Head Office.

Conclusion

AI’s impact on defence is being – and will continue to be – felt in individual steps that address a potentially wide range of issues, including how machines can be better maintained, an enhanced understanding of the causes of training injuries, and improved prediction of operational practices of potential adversaries.

It will thus consist of a set of bottom-up efforts but the new umbrella organisation, from a top-down perspective, could stimulate enhanced use of AI, coordinate efforts across the sector, and provide guidance and regulation.

It remains to be seen how its areas of interest, organisational structures and overall mission compare with those of its slightly older US cousin. So far, the UK government has decided it wants such a centre, but has not yet revealed how it should be structured and operate. One possibility, of course, is that it announced the centre without thinking through many of the details.

Perish the thought!

The views expressed in this Commentary are the author’s, and do not represent those of RUSI or any other institution.