MIT AI predicting technology improvement rates
By Ian Page
https://www.sciencedirect.com/science/article/pii/S0048733321000950
The paper has a good summary
"Highlights
- A comprehensive granular account of the rate of technological change.
- Predicted improvement rates for 1757 domains containing 97.2% of US patent system.
- Improvement rates are not uniformly distributed across and within categories.
- Fastest improving domains are predominantly software-related.
- Online system for searching domains and improvement rates based on keywords."
My main concern was that it's measuring technology improvement rates, presumably some measure of performance, rather than some measure of impact.
My thought was that making excel go 200% faster, has little impact on the world while improving the extraction rate of copper from ores from 0.3% to 0.33% increases the amount of the critical material copper by 10% and has major impacts on the market price and usage.
This leads to concluding that patents in software (where it's easy but may well be less impactful ) are the place to put your effort.
This reinforces the US/China distinction where china seems to be focussed on the basic components of the supply chain of real products, and the US is focussed on services and virtual products.
Improving supply chains is really hard but has long term strategic advantages as its hard to copy, involving huge investment and a lot of craft (non-patent) innovation. Software is surprisingly easy to copy in many cases since everything about it, is there in the code.
Comments
Post a Comment