This week I went along to the popular London Geek Girl Meetup monthly breakfast which provided me with some delicious fruit, great company and an early morning dive into machine learning.
Geek Girl Meetup is a Swedish initiative, founded in 2008, which was set up to address the phenomenon of the short ladies’ toilet queue at tech conferences, through monthly events and an annual ‘unconference’. This breakfast meetup, ‘The Magic of Machine Learning’, was informal, inclusive and included 4 speakers – all speaking for only 5-10 minutes each – who merely scratched the surface of their subjects, but opened up my imagination and curiosity to the world of machine learning.
Machine learning is a branch of computer science which deals with the theory and application of computer systems that can learn from data, as oppose to human-engineered sets of rules. It’s truly incredible stuff which, with our understanding of the field increasing rapidly, is changing the way our technology performs. You know when Amazon recommends that book you’ve been thinking about buying, or how your email filters out that spam that you really didn’t want to read…that’s thanks to machine learning.
The meetup was hosted at the offices of Swiftkey, a London Start-up which has created a ‘mind reading’ keyboard app for Android (I really hope the iOS version is released soon, as I really love the idea of using a keyboard which can predict what I’m going to type based on my previous writing activity…)
Catalina Hellett – a language engineer at Swiftkey – opened the morning with an introduction to machine learning. For me, this was much needed as, despite knowing what machine learning is and being aware of its usage and predicted growth in today’s world, I didn’t have a clue when it came to the theory behind it all. Catalina broke everything down into a simple, though not dumbed-down, explanation including Hello Kitty references and digestible charts.
We heard about 2 of the strands of machine learning (supervised and unsupervised – basically 2 different ways of setting up the learning in the first place), and the difficulties with classification (how do you explain sarcasm, for instance, to a machine? What are the rules you follow to classify speech as sarcastic…?) The intro was perfect in that it prompted further thought – it left questions unanswered which, for I’m sure a large number of the women attending, would have forced us to go find out more.
She was followed by Anna Alfut, UX Designer at Swiftkey, who put forward the strong case (through some beautiful slides!) that everything built needs to keep the user in mind. We went straight from the scientific theory to talking about an end product, reminding us that to make machine learning beneficial in a commercial sense, we must always go back to what purpose the product serves for the consumer.
Chloe Hajnal-Cereb from EDITD gave us a quick-fire case study of how machine learning is employed in this fashion retail startup. It seems machine learning really does have the power to overhaul entire industries, which made it all the more valuable and intriguing to be focusing on the subject right now.
The final talk was from Mital Kinderkhedia of UCL, a machine learning research student embarking on her PHD. She spoke to us about a more complex level of machine learning called ‘Deep Learning’, which moves the topic closer to Artificial Intelligence. It uses a set of algorithms (as opposed to just one in particular, selected for the job at hand) to perform more complicated tasks such as recognising an image. At first I wondered how hard it could be for a computer to, for example, recognise a picture of a car, but what Mital explained is that the computer only sees pixels and colours and which coloured pixels are next to which other coloured pixels. A computer would have to be programmed on several layers – to recognise colour, to recognise a collection of coloured pixels forming a line, to recognise a line forming a circle, to recognise that a circle with lines inside is a wheel, to recognise that this particular type of wheel is a car wheel…the list goes on. I really do have a newfound respect for the face-recognition feature of my iPhone camera.
The breakfast ended with a room full of energised women (and a few men!) chatting machine learning, Artificial Intelligence, upcoming technology events, collaborative opportunities and the tasks for the day ahead (for a night owl like me, it already felt like lunch time…) It was an effortless morning full of inspiration and education – which was basically free as the £5 ticket paid for the food and coffee – so I will be looking out for the next Geek Girl Breakfast Meetup with much anticipation.
(I’ve also bought a ticket to their annual conference – this year it’s ‘Ubiquitous Technology’ – which you can find out more about here)
Science Scene London 