Apparently, it’s that time again. I just gave my second invited keynote at a conference at Charité Berlin. It was really fun.
The audience were dentists – academic dentists. I confess that I struggled to understand why they thought I would be a good fit for their conference. My previous keynote was at the BIH Digital Health Forum – a much more obviously appropriate audience. But, perhaps strangely, the fit was very good.
The invitation came about through some consulting work that I did last year. Apparently I was very helpful and thus my reputation spread. That and the PyData talk I gave in 2018 sealed the deal – from their side. On my side, I hesitated a little at first but, in the end, I couldn’t resist the challenge of communicating something useful about AI to dentists.
There were actually two keynote talks. And both of our timeslots were somewhat reduced from the original plan. It was a bit of a squeeze, but we managed to deliver to the new time pretty well. More importantly, I think they did a very good job of selecting two speakers who complemented one another on such difficult topics.
The first speaker was Prof. Andreas Keller, who holds the Chair for Clinical Bioinformatics at Saarland University. Before the meeting, I found it fascinating to look through his CV. It was like looking at a different life path which I could have followed. We have very similar interests and backgrounds.
In practical terms our talks took a natural split between inference, or classification, from data – and the application to diagnostics – and prediction models. I took the latter.
Given that I had no major academic or networking goals from the meeting, I wanted to play with my talk format a little. I’ve done this quite a bit in the past, not always with the best results. My learning, over the years, has been that people have expectations and if you don’t match those expectations then you need to use strong signals to help them to navigate the talk. I explained to the audience the typical strategy of a Ted Talk – the increase of excitement following an exponential curve. I told them that the typical story of AI follows a similar curve. But now I was going to start with the Singularity and work backwards, instead, finishing with what they can do today to prepare for this future.
I think that things went pretty well. I explained the Singularity as a nice anecdote which people can tell over the dinner table. I worked back through generalised AI, explaining the difference between this and statistical approaches. I made the point that the genius of DNNs is in the ability of most engineers to apply them. This is considerably cheaper, in both time and money, than alternative approaches requiring a joint PhD in Statistics and the domain of application. We then looked at near future applications: Autonomous vehicles, Personalised medicine and Decision support systems. I discussed both the art of the possible and the likely limitations. Then I ended with a discussion of where academic dentists today should focus their efforts in order not to get sidelined during this revolution.
This year will be the first year, following three in a row, in which I will not be speaking at PyData Berlin. I’m quite sad to miss it, but I thought I had enough on my plate already. So far, I have the distinction of (I suspect) being the only person to be invited three years in a row and each time I have talked about anything but Python. They must really like me to allow me to keep coming back despite this.