Machine Learning, Defense Innovation and the British Army
A discussion of machine learning, its applications with respect to the British Army and an examination of the organizational structures required to support Research and Development.
(767 Words)
The Challenge
Like most contemporary armies – the British Army must grapple with the challenges and opportunities posed by the application of machine learning. Namely, how to ensure that the Army remains an effective and relevant tool of UK foreign policy for the foreseeable future. With respect to machine learning – the primary (publicly disclosed) opportunities relate to data analytics. The British Defense establishment collects far more data than it can process effectively and any improvements in this capability will in turn facilitate the Army’s ability to “Protect the UK, prevent conflict, deal with disaster and fight the Nation’s enemies”.
This challenge is particularly acute in the United Kingdom (vs. the USA), where as an organization it must ‘do less with more’ in order to ensure Britain can maintain any semblance of a claim to ‘Great Power’ status. Beyond absolute effectiveness, the British army must eke out operational efficiencies wherever humanly possible in order to maintain relevance on the global stage. Operating with roughly 1/11th the budget, it simply cannot tolerate waste that might be acceptable in militaries of geo-political competitors. This is magnified by Britain’s historic and geographic pre-disposition to expeditionary warfare. Conflict in the British Isles is unlikely and so to be an effective instrument of foreign policy the organization must be capable of moving/sustaining resources across vast distances. They must do so without the vast resources afforded to the American armed forces or the ‘home-field advantage’ enjoyed by Eurasian powers such as China or Russia.
Defense Innovation
Perhaps the primary issue that the British military faces in both the near and medium term is lack of access to talent. It struggles to identify and pay talent which is drawn instead to the private sector. In order to mitigate this the Ministry of Defense has undertaken a number of initiatives aimed at building out these capabilities. For instance, the Defense Science and Technology Laboratory has launched a series of prizes aimed at specific challenges posed by AI. To date, prizes have been awarded for vehicle recognition programs and information classification programs related to news media. Both aim to address the vast quantities of data possessed by the MoD and transform them into a useable format. Another example of this is J-Hub an incubator designed to finance start-ups and projects with domain expertise relevant to the army. J-Hub also provides a fast track through the MoD’s procurement process.
In the medium term – much of the UK defense establishment’s progress with respect to machine analytics will be financed by the IRIS fund (est. 2016). The IRIS fund was established and endowed with £800mm to support disruptive technologies that the MoD believes will not be adequately supported by the private sector. The fund has a 10-year lifetime and complements the MoD’s £1.5bn annual science budget. The precise portions specifically allocated to data analytics are unclear but it is reasonable to assume a substantial portion will be directed to this.
Other approaches to Machine Learning?
Historically, the MOD has had difficulty competing for top talent and is perceived by many to be overly slow and bureaucratic. To some degree this is an inevitable consequence of Government bureaucracy which cannot afford to take risky ‘big-bets’ in the way that the private sector can. For example, something similar to Google’s acquisition of Deep Mind could not occur if it was underwritten by taxpayer funds. The public would not accept such expenditures for something so abstract even if it did provide the requisite human capital. Having said this, there are probably areas where additional efficiencies can be sought. For example, it seems reasonable that J-Hub and the challenge funds established by the Defense Science and Technology Laboratory could all be rolled into the IRIS Fund. This would have the effect of ensuring common strategic goals for technological innovation as well as streamlining other administrative overhead. There seems little marginal benefit to running many of these organizations independently.
Another step that could be taken by the MoD would be to deepen defense partnerships in the context of the EU. The British Army already has links established with the anglophone nations of the ‘Five-Eyes’ but more could be done enhance research with nations such as France and Germany – providing this is strategically acceptable.
Outstanding Questions
The principal issue that occurs to me in this essay is: “to what extent does it seem prudent for Governments to partner with the private sector on issues critical to national security? Are they reliable?”. It’s clearly extremely difficult for nations other than the United States to maintain the vast R&D capabilities required ‘in-house’. How should smaller nation’s tackle this issue?
Bibliography
Burgess, M. (2017, April 2). UK military lab launches £40,000 machine learning prize. Retrieved from Wired.co.uk: https://www.wired.co.uk/article/dstl-mod-data-science-challenge-2017
Cummings, M. L. (2017). Artificial Intelligence and the Future of Warfare. London: Chatham House.
Jones, S. (2016, August 12). MoD sets up £800m fund to encourage weapons innovation. Retrieved from Financial Times: https://www.ft.com/content/56d82af4-5fd3-11e6-b38c-7b39cbb1138a
Ministry of Defense. (2018, November 07). Army: What We Do. Retrieved from https://www.army.mod.uk/what-we-do/
Ministry of Defense. (2018, March 15). jHub: What We Do. Retrieved from https://www.gov.uk/government/news/jhub-what-we-do
Thank you for sharing this post! I do not know much about military given my background, but machine learning and military defense is an interesting topic. Given UK or any other countries do not have budgets like US or China, I imagine building up machine learning capabilities would be crucial. At the same time, partnership with private sectors may be difficult given the confidential issue. It might make sense for the government to work with universities on machine learning as part of research topic to use their capabilities as well as non-profit identity. Another idea is to use best practice by other nations such as Israel. Based on Alex’s presentation during flag day, I remember that Israel has a strong technology industry, which works closely with the military. The model they are using may be applicable for UK or other countries which do not have enough budget, yet struggling to leverage machine learning potential.
I think this is a great article that shows the predicament of many nations as they seek ways to improve their military with limited funds. Additionally, I strongly agree that as time progresses the talent of individuals seeking to join the military is diminishing, thus the governments need to find ways to improve their armed forces when a big portion of the capable population is not interested in joining its ranks. My question with developing machine learning for this situation is that if the UK modifies their force due to inaccurate machine learning conclusions, the results of this will be not only very long lasting, but can also place their current military force in mortal danger, thus it is of vital national security to ensure that the machine learning results are accurate before any action is taken.
Thanks for such a thought-provoking article. I particularly enjoyed how you are bringing up opportunities that bridge the knowledge gap between the public and private sectors. When it comes to investing in and adopting new technologies like machine learning, government organizations (e.g., the British Army) should create a separate division that is able to experiment and test in a slow and controlled environment. This way, the process aligns with the government’s pace of adopting new ideas, and at the same time collect enough data points to improve on the machine learning applications before broadly implementing to other system-wide departments.