After two weeks of stay at home lockdown the garden is much tidier and we are getting rather sated with social media videos. But there are still some goings on to amuse us despite or, in some cases, because of coronavirus. The news item that made me gasp the most was an interview with Professor Neil Ferguson of Imperial College, the author of the flu virus model that has been tweaked for coronavirus and on which the government is placing so much store*. In previous epidemics ministers have called on several universities to collaborate and help work out the best responses. Not surprisingly Ferguson's rivals aren't happy. Part of their problem is that they cannot tell how Ferguson's model works. It consists of several thousand lines of dense computer code, with no description of which bits of code do what. (Which is what made me gasp in amazement). Ferguson agreed this is a problem: “For me the code is not a mess, but it’s all in my head, completely undocumented. Nobody would be able to use it . . . and I don’t have the bandwidth to support individual users.” The government must be hoping Ferguson doesn't succumb to coronavirus....
Ferguson plans to put that right by working with computer gamers including John Carmack to publish the entire program as an interactive website. Now gamers are very good at slick interfaces so there's a lot of logic in that. As Carmack is the lead programmer of the computer game Doom (I know, you couldn't make that up) we'll all soon be able to model the world as we know it being wiped out. I don't know if they'll come back as zombies.
But several thousand lines of code with no back up documentation?? So I take it that Ferguson doesn't work to ISO 9001 let alone have TickIT** certification, which is what any reputable company working in software development would have. I know universities live outside such requirements but it didn't occur to me that they would ignore them completely. During my career I worked with and sometimes had responsibility for teams working on software for railway signalling control, railway track-train interaction (i.e. does the train stay on the track....), railway timetabling, radiation physics, modelling of the dispersion of hazardous materials and economic modelling of nuclear and other energy sources. I've just had some fun reminding myself about some of this stuff***. All of this means I know a little (a very little) about how such work is planned and controlled. I often used to say "I'm glad I didn't have to do all this quality assurance stuff when I did real work rather than going around asking dumb questions all day" because it can be a pain in the backside. But very early in my career, in the chemical industry, long before QA had been invented, I had to try to pick up some computer programs written on another site by people who had left the company. It was nigh on impossible (we ended up rewriting most of it so we understood what it was doing). The chap had been an angler so even the programs were named after fish rather than what they did.One was called pike I recall. After that whenever I did any programming (never that much but there was some) I always included copious comment lines in the code to remind me or tell others what each bit did and where I'd got the formulae, algorithms and data from.
So Ferguson's comment, in the context of the weight being placed on his model, was perhaps the most gobsmacking thing I've read in yonks. Oops, shame we have to rely on him. But do we?
There's a Royal Society initiative to create a network of disease modelling groups amid concerns that the group at Imperial have a monopoly of government attention and that some competition would be a good idea. Behind this are academic and personal tensions between the country's 100 or so academics who are disease modellers, who will all know each other and be used to competing for jobs and grants and so are doing the usual thing of competing for work, or at least attention.
The Sunday Times interview noted that Ferguson had faced challenges from other modellers, which he brushed off. But this week they followed up on it. Sunetra Gupta, a professor of theoretical epidemiology at Oxford published a paper suggesting that some of Imperial's key assumptions could be wrong. Could be? Will be, at least in the detail, for sure. She reckons Imperial's modelling needs to be replicated by others. Gupta went on to say "....it seems irresponsible that we should proceed without considering alternative models." Mike Cates, Stephen Hawking's successor as Lusasian professor of maths at Cambridge and leader of the Royal Society project, joined in, damning the Imperial team with faint praise ("The Imperial team are good but these models were optimised for a different purpose which is influenza.....". Wow really? I could see the objection if it had been based on, say, STIs or something which spreads in a different way.) Anyway Cates said "with only one model you don't know which bits of it you really can trust."
Right, so we should be looking at the results of several different models quite possibly producing conflicting results? That sounds like a recipe for paralysis by analysis or total confusion. While the detail might be in his head, I accept the basis of Ferguson's model is pretty transparent and, if it doesn't follow the progression of the outbreak closely enough there are plenty of variables to tweak to make it do so. Provided he doesn't end up, like Boris, in intensive care....I hope someone else knows his login details or has an up to date copy of the code...
The story in the Sunday Times also noted, almost in passing, that Gupta had once failed to get appointed to a job at Oxford because her then boss and head of department, Professor Roy Anderson, alleged to other panel members that she had only got the job because she was having a relationship with one of the panel members. This was untrue. Gupta lodged a complaint, Anderson retracted, apologised and left Oxford, moving to Imperial College where he works with - you guessed it - Professor Neil Ferguson. This story was picked up and splashed with more drama the following day by the Daily Telegraph.
I am even more gobsmacked. Though not, as you'll have noticed, lost for words....
The government is placing great store on scientific advice. Actually it's placing a lot of faith in some particular scientists. I hope they've chosen the right ones. Well, it might make a good film one day, I suppose.
* Neil Ferguson's interview was in the Sunday Times on 29 March. The follow up story, Modellers behind lockdown "need competition", appeared on 5 April.
** as usual Wikipedia is a good start if you want to know about TickIT; see
https://en.wikipedia.org/wiki/TickIT
*** The "Answers" software service is now run by Jacobs. It enables customers to do all sorts of modelling of nuclear applications (reactor physics, shielding and criticality), other applications such as oil well logging using radioactive sources and all the related nuclear data that is required. See https://www.answerssoftwareservice.com/codes.html. The code that tells you just how fast you can tank a particular train round a particular bend before it falls off (and lots of other stuff) is called VAMPIRE, which was coined from the phrase Vehicle and (something I can't remember) In a Rail Environment, when acronyms were popular. See http://www.vampire-dynamics.com/. Fascinating stuff for a nostalgic nerd in lockdown.
No comments:
Post a Comment