Last week, BashTheBug was invited to Google in London to celebrate winning the Community Award of the inaugural NIHR Let’s Get Digital competition. Myself and Helen Spiers went representing the BashTheBug community and gave a short talk about the work all the Citizen Scientists are doing to improve our understanding of antibiotic resistance in TB.
We heard from all the other winners of the Let’s Get Digital competition and also from Googlers about how Google is using machine learning throughout all its products to make them more helpful and intuitive. A simple yet astonishing demonstration of the current abilities of machine learning is provided by Quick, Draw!
This is a simple browser game where you are told an object (e.g. camel) and have 20 seconds to draw it on the screen whilst the computer tries to guess what it is you are drawing. There are times when it gets it right and you are not quite sure how as your drawing is not very good, but then it has learnt (and I assume is still learning) what people tend to draw when you ask them to draw “rain” or a “keyboard”. Of course, it isn’t perfect and sometimes you think you’ve done a very nice palm tree only for the computer to think it is something completely different.
We had already considered using the huge amounts of data generated by the Citizen Scientists (540,505 classifications at the last count) as a training set for a machine learning algorithm and so had some very interesting discussions with a few of the Googlers. (In fact if you can apply for a PhD to precisely this).
I still find it surprising there are relatively few biomedical Citizen Science projects on the Zooniverse (check out Etch a Cell) so will be writing a blog post for the NIHR to encourage more biomedical researchers to consider this exciting and fruitful approach.
