Chess Boxing and AI, anything in common?
There is a sport called Chess Boxing. Who knew.
By all accounts you step into the ring and start fighting your opponent. Three minutes later a bell sounds. Immediately, you take off your gloves and play a game of chess - right there in the ring.
Three minutes later - ding ding ding! - now you are boxing again. This happens over and over until one person is still standing. The hard part about chess boxing isn't the boxing or the chess, it’s the constant switching between fighting and thinking.
So, do Chess Boxing and Artificial Intelligence (AI) have anything in common? Well, not directly…but AI (and the interrelated technology hardware, systems integration, applications, algorithms, process, talent and data components that make up the AI stack) will help people with the cognitive burden of switching between 'fighting' and ‘thinking’ in a Defence context. It will help Defence to
Enhance understanding and information superiority: AI will accelerate the fusion of data and subsequent analysis, and the assessment and dissemination of information and intelligence.
Enable next-generation warfighting: Allowing Defence to adopt AI to protect and equip our forces to win on future battlefields, and in the information environment, where algorithm superiority will become the foundation of warfighting.
Protect Defence with robust and responsive Cyber Defence: Improving the efficiency of our defensive cyber activities.
Strengthen and modernise Defence and the Armed Forces: AI represents a 10x force multiplier and the core to transformation across all Defence functions, improving the speed and accuracy of repetitive tasks, and creating an efficient and modern Defence enterprise.
At a recent techUK industry event my team and I gave a presentation on how we are delivering against the Enabling Warfare in The Information Age Defence Digital Function Plan, by implementing the Centres of Expertise for Data and AI for Defence. At the event we outlined our overriding principles for AI in Defence:
We will identify large and complex problem sets that span TLBs and HLBs - tackling these joint, hard, cross-cutting AI challenges using a cross-functional team approach - and ensure users have access to continuously improving repositories of data, tools, standards, processes and relationships.
We will establish a common IT backbone with a data and digital foundation for decentralised AI development, testing and deployment.
We will develop strategies for deep partnerships with industry, academia, allies and partners to leverage the best skills that the UK has to offer
We will develop and reskill our people to field a leading AI workforce
We will take the leading in developing military AI ethics and safety.
One of the challenges faced in Defence in relation to AI is getting to a consistent taxonomy and a good level of understanding of the technologies and concepts. For example, we often speak about AI, Machine Learning and Autonomy, but then conflate automation with autonomous vehicles as it is still more comfortable for people to think about physical platforms and robotics.
It’s important to address the hype curve as well. The wave of success we are having with AI is due to type of machine learning – a form of ‘narrow AI’ that offers very effective techniques for speech recognition, object detection, image classification and natural language processing. We are responsibly developing and integrating these ‘narrow AI’ capabilities to help prepare us for improvements in the way machines might be capable of broadly the same intellectual tasks as humans in the future.
We are exploring the use of AI and automation on many different scenarios, for example
Supporting Intelligence Analysis – extraction of features from text, imagery and video & understanding how to store & exploit
Countering Fake News – enabling the rapid identification & defeat of fake news
Removing people from dull and repetitive processes through business process automation or using ML to replace routine tasks
Cyber Defence & Network Ops –learning from past vulnerabilities and observe anomalous behaviour to detect and respond to unknown threats.
Two specific examples include:
Automating Statistical Reporting – The Reproducible Analytical Pipeline (RAP)
Producing official statistics for publications is a key function of many teams across Government. It’s a time consuming and meticulous process to ensure that statistics are accurate and timely.
The Price Indices Team in Defence Economics provides the latest inflation rates and forecasts, assesses and advises on application of Variation of Price (VOP) and other price uplift mechanisms. They had developed the Indigo web tool, which displays historic and forecast rates for around 180 price indices (the latest inflation rates and forecasts) and is used by commercial and cost assurance teams across MOD. The tool is currently updated quarterly through a mainly manual process (with some visual basic) and we wanted to address the following challenges:
Teams spend too much time moving data between various software
Teams can’t easily reproduce their most recent publication’s stats
Teams could be freed up from copying and pasting to add value by focussing on complex tasks such as the interpretation of the statistics and communicating the implications of these findings to others.
With open source software becoming more widely used, there’s now a range of tools and techniques that can be used to reduce the production time of such processes, whilst maintaining and improving the quality of the publication. We leveraged learning from developing the Reproducible Analytical Pipeline (RAP) in other government departments, to implement an alternative production methodology for automating the bulk of steps involved in creating the Defence Price Indices statistical report.
By providing data science, DataOps and DevOps support to our customer, we’ve helped them to:
replace monotonous processes;
improve the accuracy and timeliness of the official statistics;
free up analyst time to focus on the interpretation of results;
build a process that is fully transparent, auditable and verifiable – reducing risk and improving quality.
The project is now continuing into Live, with further support to the customer as they run the RAP for the next quarterly report. Given there are a significant number of official National Statistic products in Defence, the potentially savings could be quite large.
Machine Learning Supporting A Cyber & Network Analyst
This was a project that started with @Ben Parish's innovation team, who took on the challenge of how MOD cyber vulnerability analysts and security teams can use new advanced analytics methods to identify events of security interest in increasingly complex and dense network data - volumes that are now difficult for a human to manage. The goal was to surface events for specialist review through automated tools, allowing human operators to better use their time, focussing their efforts on specific risks.
Now we know, there are many highly sophisticated cyber security products on the market which deliver new capability, but Defence's challenge is not only the cost of purchasing and integrating but also the issue of procuring 'black box' solutions, which means that the MOD may not understand how the solution generates its results, something that is key as we build our trust in such solutions, hence building our own solution at this stage.
The project started as an Alpha project in early 2019 to test the hypothesis that Machine Learning and Data Science can assist in making cyber security operations quicker and more effective. Building on the initial validation of the hypothesis, and working with the Service Delivery & Operations Defensive Cyber Operations Delivery Team, this progressed into a Beta phase to prove the methods against a representative MOD data set. This successfully accelerated progress in this exciting area and has led to the transition of an initial capability into Live Service.
For MoD this is a great example of @Ben Parish's innovation team taking on an operational challenge, leveraging emerging technology and working closely with my Data & AI CoE team and our operational colleagues to work differently. The project is now being transitioned into live, moving from the innovation team to a local data science team will now operate as a satellite team of the Data & AI Centres of Expertise (COE) which will share best practice and our learning from this project across defence
This project has already delivered benefit to the personnel within the Cyber Security Operations Capability (CSOC), by identifying data science techniques that will be of benefit in future whilst already revealing previously undiscovered knowledge contained within the data we already hold.
So, as you can see, we have outlined the overriding principles for AI in defence and we have started delivering a number of projects in to live service that will start to free up time for our people to do more 'thinking' and focussing on value added activities rather than 'fighting' through volume / repetitive tasks.