danconnolly.co.uk

Bayesian inference

January 26, 2015

The Antikythera Mechanism is a 2000 year old computer; a mechanical calendar that could calculate moon position and phase, solar eclipses, the dates of Olympiads and more. It was discovered in 1900 (it took us a hundred years to work out what it was for) and it’s the only such device we have from antiquity.

So did we luck out and happen upon the only example of such an advanced piece of technology? Or is it more likely that this is just one of many computers? Tyler Cowen talks about what Bayesian inference tells us over at Marginal Revolution:

So what to infer?  The first option is that this device was a true outlier, standing sui generis above its time.  Cardiff University professor Michael Edmunds “described the device as “just extraordinary, the only thing of its kind””.

As an artifact that is true, but is that so likely in terms of broader history?  It is pure luck that we fished this thing out of the Mediterranean in 1901.  (By the way, further dives are planned to search for more parts of it.)  The alternative possibility is that antiquity had many more such exotic devices, which have remained unreported, at least in the manuscripts which have come down to us.  That would imply, essentially, that we don’t have a very good idea of what antiquity was like.  In my view that is the more rational Bayesian conclusion.  It is more likely than thinking that we just lucked out to find this one unique, incredible device.  To put it another way, if you found some organic life on a traveling comet, you ought to conclude there is more of that life, or something related, somewhere else.