Friday 16 October 2015

Alan Turing (and how he might have helped with my homework)

I am re-reading Andrew Hodges' book 'Alan Turing: The Enigma'.

In 1974 I was working with another planning student, developing a practical trial of a computer model called the Decision Optimising Technique (DOT). We were mentored by the model's creator, a PhD post-graduate student named Stan Openshaw. According to Wikipedia, he, like Turing, was a believer in human-competitive machine intelligence (and I gather became something of a world-class authority on the subject). There had been considerable work on the modelling of systems behaviour but, beyond critical path and cost benefit analysis, decision-making had fallen behind. Our work aimed to take forward the AIDA model (Analysis of Interconnected Decision Areas), which had a number of flaws - in particular its incremental nature and reliance on over-arching value judgements.

The Newcastle computer occupied an entire basement two floors underground. It featured ranks of steel cabinets with memory provided by spinning magnetic drums. Access was via remote keyboard terminals, in our case using stacks of punched cards which were fed into the machine on batch at night.

Even though this leviathan IBM computer could be linked with its siblings at Edinburgh and Imperial College in London to increase processing capacity, we found that all but the smallest real-world problems were too complex for the runs to be completed. While my colleague wrestled with the serious maths and machine coding, I found more and more of my energy being devoted to devising ways of limiting and reformatting problems to reduce the scale of calculation the computer was being required to address. Initial attempts to use 'IF' statements or positively/negatively link decisions by use of 'bars' (as in the AIDA model) were not enough, and amongst other tricks we hit on the idea of assigning token numerical values to individual decisions, allowing us to use constraint equations as a substitute.

It is fascinating to learn that Turing was struggling with exactly the same problems when he was trying to break the Enigma code on his 'Colossus' at Bletchley Park, and after the war when he was developing his 'Automatic Computing Engine' (ACE...arguably the first electronically-stored-program computer). Like us, much of his time was spent devising ways to reduce the iterations of the machine, and he too realised by giving instructions bogus numerical values he could use equations as a substitute for 'IF' statements. I suppose similar problems spawn similar solutions, but I wish we had known about his work then.

It is odd to realise that we were working only twenty years or so later; that the computer we were using, with its punched cards and steel racks, was ACE's direct successor; and that the bog standard PC I am writing this on would probably be capable of running DOT in minutes, not hours or days.

2 comments:

  1. And now we walk around with smartphones that can do all sorts of wonderful things! Who knew how technology would evolve and how quickly. I'll have to look for that book, sweetpea, since I've seen the movie. (I know, backwards, but what can I say? I'm an American and that's how we roll.) ;) Good to see you again, sir! xoxo

    ReplyDelete
    Replies
    1. And to hear from you, Savvy! I must warn you, that book is a hefty read; full of maths (not my strong point), so I keep having to go back and read paragraphs again. The movie was good though!

      Delete