Many people have told us that
more and better use of new internet technology could have massive benefits
in preventing climate extinction, at much lower costs than some of the giant
Christmas tree climate proposals we have seen.
That is absolutely true, and
it is a matter of life or death, but it is not happening because of
institutional problems.
When I worked at NSF, I got
to see how those kinds of problems can work, first hand. For example… to really
prevent climate extinction, I have argued for a 5 point approach, where point
one is reduced net CO2 from making electricity. I have seen how RD&D money can be wasted in that area, and also how it can be massively transformative.
SOME funding agents would
look for those projects which can help, which have the lowest level of risk. Some
would focus on the big targets for how AI could make the maximum contribution,
regardless of risk. Our big problem today is the diversion of funding to
less relevant, lower risk projects where the AI people don’t have to learn much
new, as if they were auditioning for jobs as janitors, but the core needs of
climate change are not addressed as well as they could be, by orders of
magnitude.
This kind of problem came to
me very vividly in 1993 or 1994, when I was a reviewer for the interagency
defense conversion research program:
SR-1994-Defense-Conversion-Transition-and-the-Industrial-Base-Report-of-the-1993-AUSA-ADPA-Joint-Panels.pdf.
The people who built that program created a list of review criteria to use in
deciding which offer to fund. In that program, I once saw a choice between two
offers:
(A(A) A
near certain min, delivering $1 million worth of net new benefits
(B(B ) A 50% serious but risk program, worth billions
if it succeeded
The review criteria we were given strongly favored A. Most
workers would strongly prefer A, which would give them secure income and a
secure future. Many bottom up efforts reflect those preferences for A. (This reminds me of Schumpeter's prediction about how hard it would be for the US to maintain the spirit of honorable competition in general, so essential to growth in productivity.)
I have seen many massive consortia funding stuff like AI for
good, to help with climate change and such, and most of the people end up in
type A consortia. I did see one great group in Canada, with leadership from Michael
Barnard, which focuses on what we really need to solve the real, important
climate problems, but big public funding and attention is mainly on other
activities which become like Christmas Tree programs.
There are a few big flagship programs in special places,
like Dubai, where one on one personal connections do focus on big goals, but
without the kind of general competition we need to mobilize the whole range of
the best technology we need.
What WOULD work?
After I ran the NSF electric power area for a few years
(mainly 2010-2014), I summarized what we really learned, what would really
work, in a paper jointly funded by power people in France, Chile and Brazil: www.werbos.com/E/GridIOT.pdf .
Above all, project selection was open and competitive, and
the criteria were well focused on the goals. Maximum serious potential benefit
to what really matters was operationalized. I even remember part of instructing
review panels: “Do not focus on what the worst case risk to the project is.
[Worse case risk to earth climate is a very different kind of risk, since we only
have one earth.] For each project, ask how big the loss might be if we don’t fund
it, if it would have succeeded.”
The
strange thing is, I learned at NSF that the projects which SOUND like low risk are
usually MORE likely to fail, to end up as a waste of money. That is because
they take approaches others have already tried, and do not invoke the kinds of
new directions necessary to a big change.
So which do you want to buy: one big antiquated flagship,
1000 janitors, or 100 tough pioneers of whom 50 (or even just 10) produce
something which totally changes the game? With climate, we KNOW where real game
changers ARE doable, but we need a radical change in funding cultures to
actually make it happen.
Until about 2013, when Lamar Smith began his remaking of
NSF, NSF was maybe 25% efficient in doing this kind of stuff, which made it the
number one wonder of the world for what it did. Proposals to totally avoid risk
in NSF itself would have led to 0% productivity, through uniform type B funding
or flagships by friends/graft. Where it will go now, post Smith, is hard to
predict, but the world needs something like the classic NSF to do better on
worst case climate risk. That included many elements which need to be
reinvented, starting with Vannevar Bush’s initial manifesto and updated with
new computer systems.
And, of course, never ever to forget the larger goals.