I think the push to autonomous medical devices say radiology interpretation
machines is misguided. It is much better to provide better computer tools for
radiologists. Microscopes can "see" better than any medical researcher
for example.
The use of position analysis tools for world class chess players is
an example. World class players can now beat chess programs because
there is much more financial support for world class chess players.
In the 1990s IBM chess computer program development group had a huge budget
that included chess experts that were very good grand masters but not as
good as Kasperov who did not have any financial support. Fabiano Caruana
is probably the best current chess player at using chess position analysis
computer programs to make game tree discoveries and then win chess matches.
He has used that skill to distance himself from other world class
challengers. He won the qualifying round and will challenge Magnus
Carlsen for the world championship in the fall.
He is part of the Saint Louis Chess club that according to the Boston Globe has
support from "Rex Sinquefield's seemingly endless financial support."
(
The Boston Globe)
The best humans can now beat any chess program. There probably won't be any
matches between chess programs and human players because it is in the economic
interest of players to lose to programs. The situation since the 1996 Kasperov
match is documented in my falsification of AI paper
(arXiv:1704.08111v2 [cs.AI] - new version). It includes a reference to Kasperov's
book in which he accuses IBM chess group of bullying tactics.
The same situation occurs in medical specialties. For example as medicine
advances, there are more cases where rare diseases need to be treated. The FDA
even has a special category called the Orphan Drug Program and tax credit to
encourage development of treatments for such diseases. It is a lot better to give
say radiologists better computer tools that to continually need to update and
distinguish rare diseases in autonomous diagnostic programs.
