Bernard Silverman outlined his long involvement with IMS in his speech after the delicious banquet hosted by LSE at the London meeting:
I’m very honoured to be invited here. It’s fantastic that the IMS meeting is in London, and I really hope that people are enjoying their time here.
My own involvement with IMS goes back nearly 50 years, so I thought I’d reflect a bit on that. Early in my career I can remember meeting some of the founders of our field. For instance, I went to a conference in Chapel Hill in honour of Hoeffding, and I was introduced to Neyman on his 85th birthday (which was in 1979). When I was a student in Cambridge in the mid 70s, we had a visit from David Blackwell and I was deputed to take him and his daughter punting. About the same time we had a visit from Grace Wahba, who has been a role model and a friend, if I may say, ever since. And while I’m thinking about people whose work has been hugely influential in our area, I should mention David Cox, who sadly died recently, and C.R. Rao who is still alive at the age of 101 (I’ve never met Rao, but I did meet Kolmogorov at a strange dinner in Moscow in the mid 80s). I often used to meet David Cox in the supermarket in Oxford and on the train to meetings in London; the last time I saw him was in March 2020, when we had a long talk in the street and then he apologised and said that he had to go because he would be late for work in the office; he was 95.
Apart from the people, it’s also interesting to recall what I used to do when I was a research student in the mid 70s. I used to spend quite a bit of time in the library simply reading the journals. And by that, I mean when a new issue of Annals of Statistics or JRSS-B came along, I would actually read every paper in it. And a lot of back issues of the old Annals of Mathematical Statistics. The RSS would have a Read Paper every month and a whole group of us would travel to London, regardless of the topic, and listen to (and often take part in) the discussion. David Cox, then at Imperial, would almost always stand up and make a pithy and apposite point. The RSS read papers were a sort of running survey of all the leading areas of research and hence it was an excellent way of keeping up and building a community of friends and colleagues.
Maybe it’s inevitable, as our field has grown and developed, that it’s unlikely a student could have the privilege of such a wide view nowadays. Electronic journals are wonderful, but there’s nothing like an old-fashioned paper volume to browse through (though my browsing days finished well before electronic publishing became the standard!). Nor would it be so easy to get to know so many of the pivotal people because there are so many of them and they are so disparate. Societies like IMS and RSS were always important venues for interaction and publication, but they do far more than that, because they are actually, in my view, the glue that holds scientific activity together.
In the 70s, I wrote a paper called, “Density Estimation: are theoretical results useful in practice?” My friend Professor Paul Deheuvels from Paris said, “Yes: they are useful for writing papers.” At the time, statistics was only considered a useful subject in certain scientific areas. You could do biology without knowing any maths at all. And of course the famous physicist Ernest Rutherford is reputed to have said, “If your experiment needs statistics, you ought to have done a better experiment.” And as for what was actually possible for statistics, my Master’s course included learning how to use a hand calculator to do analysis of variance—scientific programmable calculators didn’t really exist at the time. [Editor’s note: Bernard slightly later co-designed the first programmable pocket calculator.] Now, we have the whole fields of machine learning, data science, genomics, and so on. Both mathematical statistics and probability are fundamental in these areas.
So here we are, working in an ever-expanding field (or fields), and no longer can anyone seriously think that our subject is only useful “for writing papers”. So, I wanted to reflect on what role the IMS plays.
The IMS was created in order to create a “home” for mathematical statistics. IMS has a distinctive voice, which I would see as supporting the fundamental understanding of probability and statistics—for their own sake, but also as underpinning so many other disciplines and activities. Data science and machine learning can too easily degenerate into a mess of ad hoc tools if we don’t get to grips with actually understanding what is likely to work and why. No amount of computer simulation can beat probability results that are actually established mathematically. We should be proud and confident about all the hard and rigorous thinking that goes into our discipline, and the IMS—its people, its journals, its meetings, its whole philosophy—represents that. Many people think we don’t need learned societies at all, but long experience shows that societies define and support their disciplines and hence, even those who don’t realise it are benefiting from them.
Think about a learned society as something that puts a wall around a discipline to protect and define it, but it needs to be a flexible wall with lots of space for air and ideas to flow in and out!
There’s been tension between probability and statistics within IMS for decades. This is unsurprising, because as disciplines become wider, the ability to understand any more your own part increases. But we should really resist that. No longer can anyone individually get something from all the journals and all the talks—but the organisation as a whole does have that corporate understanding, and we should all support it.
As I said, the IMS is the glue that holds all this together. By coming to a broad-ranging meeting like this one, you’ve all committed yourself to that broad view. Thank you. Keep it up.
In closing, I’d like to propose a toast: to probability and statistics, and to the IMS!