The box is never empty - the story of our AI Hackathon

You walk into a room and find a small white box. Carefully, you open the lid. What do you see?

Some people see a diamond. Some people see an apple. Some people see a frog. What you see doesn’t matter: the point is, the box is never empty.

I heard this about somebody who trains comedians to do improv. Her students are terrified that when they stand up in front of an audience they draw a total blank. 

But what she teaches them - using exercises like this - is that you always come up with something. The box is never empty

Well, this is how I felt as we started to plan the AI Hackathon we ran a few days ago. I wasn’t sure what we’d get at the end of it, but I was sure if you be good. And it was fantastic.

The whole journey has been a really interesting learning curve for the team spanning over four months of planning, preparation and problem solving: everything from securing Defence Accelerator funding for a follow on Open Call, through to getting approval and ordering the catering (I think the funding was easier)

We started the journey in Aug 2017 when we applied to the Defence Innovation Fund for funding to run an AI Hackathon. From the outset, we wanted the event to be a MoD wide event, rather than just an ISS one. This proved to be a crucial decision as it was key that we had challenges from across defence and the associated data sets to give the hackers something to work with, more of which in a minute.

The list of things to arrange for the event, seemed to initially grow with each passing week. These included things such as multi-party Non Disclosure Agreements, a venue (Geovation in Farringdon) and marketing & comms, but the two most critical items were the challenges and the data sets to support them. 

The first day kicked off with a couple of opening speeches around the importance of AI from MoD and ISS closely followed by the challenge owners outlining their challenge and why it was important to MoD

There four challenges were:

UK Veterans - Informing our strategy for implementing the armed forces covenant. This challenge was to combine anonymised veteran data with a range of open source data, such as the Census, Local Gov’t Housing Data and Employment data in order to look for correlations or differences in the Veteran population’s housing, employment or education levels against the general population, which can then be used to inform policy and services at a local level. MOD are inundated with information requests for this, and making the right data more easily accessible is ideal.

Navy Command - Situational Awareness using Automatic Identification System (AIS) data. The challenge was to explore techniques, trained by historic AIS data, to identify anomalies in shipping movement, allowing the Royal Navy to classify potential threat or friendly vessels as they make their way across global waters. If this could be achieved through the use of AI, then Naval resources can be redirected to solve other problems or threats and react with more agility to these insights.

Defence Intelligence - Raster Mapping From Satellite Images. Often, MoD are required to rapidly deploy solutions in order to overcome DI capability gaps in a particular situation. It is sometimes the case that the quality of mapping available to MOD either lacks detail or is out of date. It is also the case that the accurate, updated mapping is not always available in a timely manner to reflect changes to a human or natural landscape following a natural disaster or major conflict. We wanted to explore whether machine learning, particularly deep learning, could automatically create and update mapping from satellite imagery.

Information Systems & Services - Supporting The Big Decisions. ISS have recently implemented a time recording system. The aim of this tool is to help deliver greater transparency around cost and utilisation of resources, allowing ISS to make informed choices to be better at achieving customer demands. In addition to a range of MI, it is acknowledged that there is a lot to learn in the management and analysis of this data, with the objective of sustaining a dynamic, balanced delivery portfolio. How else can analytics tools and solutions such as AI and Machine Learning be exploited to ask the informed questions that ISS would not think to?

At this point, it was fascinating to watch the individuals ‘offer’ their ideas to the room and the teams form, 14 in total. Then it all went quiet and the coding started. Well OK that may be a dramatisation, but certainly the noise level dropped as the coding started.

This continued through to 8.30 in the evening, by which time the attendees had demolished nearly a hundred Pizzas and headed off to the pub for the evening.

The following day started with a quick update from the teams on their progress so for and then continued with their solution development through to 2.00pm when at which time they were interrogated by the judges, before getting their presentations ready for the 2.30pm pitching session.

Prizes were given for a variety of categories such as best presentation, most innovative idea etc, plus three solutions were recognised as those that we would want to see submitted to the Defence Accelerator Open Call where we have secured ￡400,000 to fund 3-4 projects

Here’s the Hackathon in numbers 

At the beginning: 120 attendees, 38% of attendees had never worked in Defence, 85% had never attended a Defence Hackathon. The main objectives for attendees were to meet new potential partners and improve their knowledge of Defence 

In the end: 14 teams presented - 6 teams were awarded prizes, 3024 interactions on Glisser (an interactive platform used throughout the event), 83% of attendees succeeded in meeting new potential partners, 90% left with an improved understanding of Defence Challenges, and 90% think Defence should run more events like it!

And some quotes: 'It was a fantastic event, perfectly organised' Deimos Space UK, 'It was a great event. Really appreciate the opportunity to attend' Benz, '100% pleased I did attend although it was a little off expectations' Enemy Gadgets,

Finally, I would like to thank IBM for providing all of the IBM Cloud platform for the teams to use and PA Consulting Group for supporting the organisation of the event. Finally I’d like to thank my ISS Innovation team, all of my colleagues from across defence and the participants, all of whom contributed to a fantastic event.

With such great success and collaboration, what are the next challenges for Defence to address?