[p2p-research] Building Alliances (invention vs. innovation)

[J. Andrew Rogers]

On Fri, Nov 6, 2009 at 4:49 PM, Paul D. Fernhout
<pdfernhout at kurtz-fernhout.com> wrote:
> To invent something takes enough money to stare out of the window for years,
> which is a lot to an individual, but not much in the scheme of things.
>
> To take an invention, and dominate the marketplace with it, calling it an
> innovation, that generally take a lot of capital. :-) And it seems to me,
> that is more what Andrew is talking about.


I'm actually talking about both, because in some ways they are not all
that separable in a well-functioning system. A lot of interesting
research is essentially forgotten because of incompetent (or
nonexistent) reduction to practice.

The point that you are not getting is that the various bits seem like
they could be done on the cheap, but you have to use individuals that
can do the whole thing from end to end.  Those individuals are very
rare and consequently very expensive; even in your minimum basic
income scenario, you can't force people to do the parts they don't
want to do but the work does not partition for all practical purposes.
Hell, open source software has giant implementation holes and many of
those apps have tons of volunteers and people that could legitimately
contribute if they chose to. Speaking from my own experience writing a
lot of open source code in the past, I only did the parts I wanted to
and ignored the rest.


Real software R&D doesn't produce things an ordinary programmer can
contribute to. Take, for example, sensor fusion -- a real problem --
where you merge unrelated measurements of reality into a single,
coherent model on a computer. Use cases are things like virtual
worlds, Microsoft's Photosynth, agricultural modeling in the
developing world, cloud-based relational databases, and numerous other
applications, all with severe limits based on a "simple" problem in
theoretical computer science that has been around for many decades. No
existing strategy in literature points to a solution so it will have
to be something new. So you purchase the entire body of relevant
computer science literature, analyze it, and clean-room a brand new
theoretical basis for a solution that actually works based on a
previously unknown mathematical relationship between constructs in
topology theory and join calculus with some other ideas thrown in. If
you publish it, a dozen people might read it, half of which might
understand it. Your programmer basically has to be a theoretical
expert on topology theory and join calculus to correctly implement the
solution in the context of one of the above applications. That's a
problem.

In the real world no one competent volunteers to see it all the way
through because there is a lot of slog work involved and no one cares
about the whole process. Additionally people that can solve problems
like this are exceedingly rare and valuable, so they have other things
they could be doing even if it might be less valuable to society in
some absolute sense. This is an extremely effective model, and you
rapidly get results that would be nigh impossible to get any other way
because you systematically force reduction to practice, but it is also
very expensive.


> So, Google invented a search algorithm, maybe better than others at the
> time, but probably not something others would not have invented in a few
> years. But, then they partnered with people with money, and turned that into
> a big company that centralized control of the invention is an innovative
> thing.


Google's search algorithm was not original to anyone familiar with the
literature.  What was original, and ultimately made them the success
they were, was the design of their database that allowed them to scale
the way they did.  Again, all they did was take some obscure ideas in
academic literature that had never been seriously implemented in a
real system and figured out how to put them together. Any other search
company could have done the same if they had been able to get the
right set of bright polymaths working on the problem that were
familiar with some seemingly unrelated bits of literature.

And this was for a problem where the elements of a solution are widely
understood by computer science geeks.


> But, one could imagine that the people who started Google had just published
> their idea, and further, that someway was found to adapt it, in a
> "Folding at Home" distributed computer way to use idle CPU cycles and a p2p
> meshwork. We might not need Google server farms at all to do fast searches.
> But, there is not much incentive for people to spend millions or billions
> developing that idea, even if it was better and more secure and more
> democratic, because, in a captilist society, where is the profit?


There are well-understood theoretical reasons why that won't work.
Centralization was required, not a design choice. No need to look for
the profit bogeyman when simple mathematics will do.


-- 
J. Andrew Rogers
realityminer.blogspot.com