A Loss for Machine Learning
As a machine learning researcher in the UK, I’m in a very fortunate position. A lot of the greats are based in the UK; Gharamani, Bishop, Williams, etc., while many others like Hinton, Hellers or LeCun frequently visit. Thanks to the UK’s relatively compact size, it’s usually not a problem to get to meetings if you really want to see someone give a talk. I’ve seen talks by Hinton and Bishop, talked to Gharamani and Williams, and have generally been very fortunate to know or experience a large swather of the machine learning greats. But on 14th April, we lost someone that I really should have met, yet, somehow, never did.
I’m talking, in case you hadn’t heard about it yet, of Professor David MacKay. If the name does not ring a bell, it’s because he mostly left the machine learning field about a decade ago to concentrate on his passion for sustainable energy. He made great contributions to that field, but I will not list them here, as they have been covered elsewhere. Instead, I want to concentrate on the impact his earlier machine learning research had on me when I started my career.
I first got exposed to David’s work when the supervisor for my 4th year Honours project, Amos Storkey, pointed me towards his book for a thorough introduction to the essentials of Bayesian statistics. I found it approachable and helpful, but did not think about David again until a couple of years later when my PhD supervisor pointed me to some of his early papers on integrating out nuisance parameters. Only then did I realize how fundamental an impact David’s work had on the early development of Bayesian statistics. By then he had already scaled down his research to go advise the government on sustainable energy.
Through the rest of my career, I would be aware of David’s work; people mentioning him off-handedly in a variety of different context; and I would always think that one day I would like to meet him. I only made one real attempt to bring this about, back when I was reporting for the EUSci student newspaper and tried to get an interview with him. It didn’t work out for a variety of reasons, mostly the fact that he was exceedingly busy at his government advisory job.
On 14th April, I lost any chance of meeting David MacKay in person, but his contributions to machine learning remain, as does his book on Sustainable Energy, which I intend to read as soon as I can. What also remains are his blog posts, chronicling his life and his final struggle with stomach cancer. They show a man with humour and thoughtfulness, who worked hard to improve the world, and succeeded.