The Royal Society held its first ever hackathon, themed around machine learning and Hacking Happiness, in partnership with Digital Catapult, on 16 and 17 January. This was also the first hackathon for me.
This event was part of our ongoing machine learning project, which is looking at the opportunities in the field over the next 5-10 years. What better way to show the potential of this technology than by putting it into action? Here, I will tell you about my brief experience as a hacker.
Happiness: what’s machine learning got to do with it?
The opening keynotes provided an introduction to happiness and what machine learning could do. Dr George MacKerron, Chief Technology Officer at PSYT, a company developing technology enabling wellbeing, showed us different flavours of happiness and some interesting scientific observations. Did you know that having the colour green in our environment can increase our happiness?
David Fearne, Technical Director at Arrow, illustrated how the company provides machine learning as a service. For instance, Arrow used natural language processing to analyse the social media threads that victims of abuse agreed to disclose, and helped Victim Support gain insight into when victims might have started experiencing abuse.
The power of machine learning was echoed in the keynote of Professor Marcus du Sautoy, a member of the Royal Society Machine Learning Working Group. Marcus talked about how rapidly machine learning has developed, and about the huge opportunities for multiple sectors of the UK economy: in particular manufacturing, pharmaceuticals, the legal sector, and smart cities, transport and utilities.
Skills and teams
Seizing such opportunities requires access to the right skills, and the UK is in fact facing a data skills shortage. So, it was great to see how much fresh talent there was on the top floor of the Catapult, where the hackathon was held.
Analysts, coders, designers and others pitched their ideas and how they planned to create their own solutions to the challenge. Soon they were chatting and finding common ground, and the 15 or so pitches led to the formation of seven highly collaborative teams.
I wasn’t planning to take part in a team – I cannot even code! But after a colleague at the Catapult encouraged me to do so, I decided to join team Borough Meter (MP3), who set out to develop an app to show you which London borough is most likely to make you happy.
There was a broad range of skills in the team. For example, George MacKerron was the most experienced team member. Chanuki, who pitched the idea, had spent ten years in digital design and was, like her friend Merve, doing a PhD in data science at Warwick University and the Alan Turing Institute.
The team also included people who came much more recently to machine learning and used this hackathon as a platform to hone their skills and network. For instance, Michal, a freelance journalist, started only months ago with the Andrew Ng machine learning Coursera MOOC.
A crash course in data sourcing and preparation
Joining one of the hackathon teams gave me a unique insight into the joy and challenges of being part of a group doing machine learning. I thought I could at least help find open datasets online and prepare them for analysis. However, I found out it could take an awful lot of time.
This made me realise first-hand how important it was that Digital Catapult provided participants with cleaned datasets, including closed data from Mappiness and the AI-based health app Biobeats.
The route to innovation
The hackathon was a great demonstration of innovation in action, and I was impressed by the originality and variety of ideas and the technical ability participants showed in their demos.
- Extra Wurst (MP3) explored the concept of ‘comfortable conflict’, to create Spotify playlists that both appealed to and gently challenged the music tastes of everyone in the room – they suggested the concept could be applied to burst the social media ‘filter bubbles’.
- Face Vote (MP3) proposed to study whether happy profile photos and statements could win elections, and found that machine learning algorithms couldn’t easily be fooled by photoshopped smiles.
- Instagrat (MP3) monitored users emotions during online browsing and sent users pictures of kittens – or other happiness triggers – at times of upsetting news.
- Buddh.ai (MP3) developed a chat bot which calls users everyday, uses natural language processing to analyse levels of happiness and makes suggestions for behaviour to help increase wellbeing.
- Disconnect (MP3) (or disco) inspired by a recent French law on a “right to disconnect”, designed a virtual assistant that deals with out-of-hours emails and personal tweets, provides a summary and improves upon feedback.
Taken together, innovators from across the UK and beyond made this hackathon a truly inspiring event, illustrating that science is global, and an essential part of the route to innovation – and happiness.
A tough job
Teams had only five minutes to present their ‘hacks’ to a judging panel whose combined expertise covered machine learning, happiness, psychology, and media.
So, in the end, it was also a “tough job” for judges, to quote the words of Dr Sabine Hauert, the chair of the judging panel, and a Royal Society Working Group member. The winner was Route 2 Happiness (MP3), with an app that monitors your stress levels and advises you on which route to take to keep happy – not necessarily the fastest route. The judges could see how the app would make them happy, and potentially others too. They thought that, with further improvement, the app could have even broader applications.
If you want to hear all the teams talk about their hacks, you can listen to our podcast interviews (click the team names above). Also, check out the Catapult’s Storify of the two-day hackathon.
To find out more about machine learning, check out our infographic, and the recording of our recent Science Matters event with Professor Brian Cox. And look out for our policy report, which will be launched in early 2017.