Who will process my heartbeat?
I am always a bit slow getting the latest news, mostly because I don’t read the newspaper, or watch television, or listen to the radio. This is a shameful attitude, my friends tell me, as being uninformed is the surest way to fall into obscurantism and let totalitarianism slip through the door. But I just can’t change. Breaking news bore me to death.
If I can ever be redeemed, it will be through making fire. Our house doesn’t have central heating, so we rely on firewood to keep warm. And this is where the newspapers come in. Nothing like a few handfuls of broadsheet to start the most recalcitrant fire, regardless of wind direction and humidity index. My technique involves putting the newspaper flat in front of me and extracting each sheet separately, before compressing them as hard as possible in my fist. In this process, it is unavoidable that breaking news travel from the paper through my eyes and to my brain, and I get to find out what people were so heatedly debating a week, a month or a year ago.
That’s how I discovered this morning all the fuss around Google’s newly announced ‘anti-cancer pill’. In two words, this is about a project aiming at producing some swallowable nanoparticles which — if they ever work — will track changes in people’s bodies and alert them of the presence of cancerous cells. This is of course a laudable aim (if you except various worries, summarised here, about the safety of the method, the psychological impact of false positives, etc.) My personal complaint concerns the data. The search engine has said that it would remain in the hands of the patients and health professionals, but for how long?
None of the articles I have read about the announcement has had anything to say about the actual diagnostic tool. But the technology is presented as preventive, i.e. able to predict whether bad things are about to happen in your body. In other words, the data gathered by the nanoparticles will be sent to a computer which will decide whether to ring the alarm bell or not. This is a machine learning problem. Now, anybody working in the area of Big Data will tell you that if you want to model a phenomenon well, ‘more data is better data’. So if you want to minimise false positives, make more fine-grained diagnostics, learn how the disease typically develops, you are better off gathering lots and lots of material to train and test your algorithms. Once the technology is out there, it will be hard to stop its development. And given what it promises, it will be even harder for worried patients not to accept sharing their heartbeat with a large data processing company — be it Google or someone else.
So if we want a predictive medicine based on nanotechnologies (and if it works, we most probably will), we should start thinking about alternative modes of data collection and processing.