Two Topics the Institute Should Investigate

I’d like to suggest two topics for the Institute: cheap, excellent U.S. health care and the machine learning revolution.

Recently I’ve been reading Alan Blinder’s books, spurred by his new book on monetary history, and his comments on how expensive health care reform is have again raised a lot of questions for me. The wonderful NAS study of 2013 on “Shorter Lives, Poorer Health” shows that all the other advanced countries have much better health systems than the U.S. at much lower cost, 9 percent of GDP rather than 18 percent. Why can’t the U.S. save a lot of money by adopting any of these other systems? I’ve been living in Japan and Korea for the last five years, and I got several free (for residents, not even citizenship required) annual checkups and a hip replacement operation. The health care is so good!

On AI, I was delighted recently to read a story on Ars Technica about matrix algorithms that use fewer expensive algorithms, starting from the 2x2 case of using 7, not 8 (so 4x4 uses 49, not 64). Machine learning really started to make an impact in 2013 at ImageNet (see the AlexNet paper), and then AlphaGo and AlphaFold took off. A Japanese friend suggested we were at another cusp point like the Industrial Revolution, unbelievable at first but maybe so. Machine learning creates knowledge without human effort, so my image is the graph of nonhuman energy use, starting from about 0.5 MTOE (back forever) to 14,000 in 2010. I’d like to see a better transition this time than the 60 years from 1800 where the factory workers got zero wage increase.