Wednesday, May 30, 2007

Geographic Predictors

Ok- another potentially cool idea- ready for thesis writing. What if you could "Predict the locations of the next technological innovations and breakthroughs?". It's so fascinating if you think that Sillicon valley is home to so many innovative companies (Google + Apple just to name a couple of recent hyberbolic examples of what many associate with innovation) and also some of the best weather in the country (warm, sunny, mild coastal climate).

So Why?

Are people who invent and create stuff magnetically drawn to certain creative poles ? Really what is most interesting is what are the other politcal, cultural, business and legal (to name several) conditions in certain areas. For so many thousands of years China came of with so many inventions before modernized equivalent came to fruition in other parts of the globe. Something has to be different in more than just the climate.

Some countries produce way more patents every year than others. And what if historical data (economic, educational, scientific was used to extrapolate and predict what areas of the world most closely resemble those which have in the past fostered so much growth in new ideas, inventions, innovations. Using varied forecasting techniques predictions could be made on where money should be invested years before its too late too make 'lots' of money.

Taking a broader perspective you could attempt to predict where the next dictator or bloody riot would take place. You could also predict which Universities will pump out the best graduates each year by automatically updating the most current campus information and not simply what some tradition tells employers where to hire. I realize various attempts have been made with different types of outcome predictors- but compiling this information on a global scale constantly updated would be amazing. I think I should talked to some programmers, economists, and scientists from across the fields- this would be incredible.

4 comments:

Paul said...

The 'programming' is not the difficult part, it is the easy part. The hard part is determining what the 'heuristics' are and how to weight them approrpiately to provide probabilistic prediction. Software will not 'magically' do that for you, it will crunch the numbers once you have determined the algorith to do so.
EG: When you take in economic data, population density, average level of education, climate etc.. you have to give it a weight, 'crunch' those numbers, and based on your algorithm, say the predictability for breakthroughs is low, medium, high etc...

This sort of thing already happens in a number of fields. The most ubiquotous one is... weather. A number of data points historical and present, are taken into account to predict the weather... the better the algorithm, the more accurate... This is also used in predicting where predators and murderer's exist. EG: Most criminals are within 5 miles of their home.. and various clues will strengthen or weaken that..


With sociological and economic data, it can be done. I think it's a good idea overall. If you come up with the 'heuristics' it could be sold.. Good luck!

This sort of

Paul said...

BTW: If you come up with the heuristics, I'd be happy to provide the programming for you.. that part is really easy. It's just math, 2 + 2

Marc Jorgensen said...

fascinating. thanks for the info

Marc Jorgensen said...

fascinating. thanks for the info