Downgrading software

My car recently got a software upgrade as part of a service. Little did I think that it would bring a change I didn't want to the engine's characteristics--making it rougher and rortier--as well as fix the bugs it was supposed to. Here's the real wrench: I can't get it downgraded. Even though it must be possible, there's no procedure they'll enact for doing it. It's unthinkable that you can't get hardware back to where it started...so why the problem with software?

This incident reminded me of the nervousness I often have with upgrades and got me checking our policies and practicalities for Mathematica. Yes you can downgrade if you ever feel the need (though of course I certainly hope you don't).

"I calculate, therefore I am"

That's my talk title at tomorrow's TEDx conference in the EU parliament. Just heading out to Brussels now. I'll be piecing together an eccelctic mix of Wolfram|Alpha, Mathematica and how education must urgently follow the "real" world into fully deploying computer-based math. It's a lot for 18 minutes!

"Publishing the future"

"Let the reader drive" was central to my talk on "Publishing the future" today at the Mathematica Users Conference. I compared narrative and reference books, server and clients models, applications and documents. But whatever the publication style, technology or field, the aim should be simple: to optimize the flow and use of ideas from author to reader (and do much better at it than we manage with the 350-year-old way it's usually done today).

Demonstrations is one good example of this radical publishing approach, one that's both innovative and successful--a rare combination! 

The risk of not linking risks

Why do governments, banks, insurers and others who manage our world often assume the correlation between different risks to be far smaller than it is? That the risk of one disaster is independent of another--when so many calamities (eg. Lloyds of London's near collapse to the present financial crisis) have shown they are not? 

This was the question I posed to Professor Paul Krugman after his talk at the World Knowledge Forum in South Korea from which I've just returned. He answered that several factors were at play that I'd summarise as: greed, stupidity and governments being accused of overzealousness in assuming correlation. I'd add complexity: it's much harder to model correlated risks, so people don't or can't.