Re: [asa] Building a flagellum step by step

Iain Strachan wrote:
> Rich,
> I pointed out that a lot of science was
> a case of wishful thinking, and that the more you wanted a thing to be
> true the less careful you were about the controls on your research. I
> gave the "Cold Fusion" fiasco as an example. Here a lot of money was
> wasted at Harwell labs, where I used to work, attempting to replicate
> Pons and Fleishmann's results. If P&F hadn't been so excited about the
> prospect of solving the world's energy needs maybe that money could have
> been saved.
>

In the late 80, early 90s I was seconded to a lab that mainly worked
with main frame compilers. Coming from the IBM Toronto lab where we
were working on compilers for the RISC/Power/RS6000 among other hardware
platforms, I had a very different outlook. The teams that I had been
leading in Toronto, had been working both with the hardware developers
in Austin and the originators (at least within IBM) of the RISC concepts
at IBM Research for about five years. At the lab where I was seconded
to, the consensus opinion was strongly that main frame computers with
vector extensions were the best solution for high end scientific
computing both then and in the long term, although Cray machines were
also respectable. This lab also tended to control the peer review for
internal conference papers in my area of expertise, so getting papers
accepted was sometimes hard and after a presentation at a conference all
an IBM fellow in attendance had to say was that they disagreed or had
tried a similar approach and failed and much of the rest of the audience
would sagely nod their head and write you off. Obviously my opinion was
that the RISC processors were the coming thing and I took considerable
negative comments including things like, you Toronto lab people are too
stupid to have an opinion worth listening to. In this case each side had
support from IBM Fellows. It so happened that I was at the lab that
supported the main frame vector processor approach, the day that
Lawrence Livermore switched their mainframe high end vector scientific
processor order to a network of one of the early Power machines. In the
halls that day at work the shock was almost palpable, people seemed
stunned.

I could recount at least one other major example where the internal
consensus was wrong and the upstart(s) eventually were vindicated, even
though it took ten plus years.

Just to keep the record straight and not present a too rosy a picture I
blew it a good number of times. I'm not sure most of the list is
interested in the details but give one example below.

So my points are:

The consensus can be wrong and the turn about abrupt, unpleasant, loss
of funding and in some cases career ending for those on the wrong side
if and when a turn about occurs.

If the potential reward is big enough (eg fusion energy), some times it
pays to spend money on what to some people appears foolish, wrong,
misguided, low probability of success. And some (often?) times the
gamble doesn't pay off but it is often hard to tell ahead of time as I
suspect Randy would agree.

At this point in time I am not aware that Wm Dembski has gotten one of
the really top statisticians to even give a qualified endorsement of his
ideas and I am not sure he ever will get such an endorsement. However,
having been in the position of presenting ideas to a hostile group of
IBM Fellows it can be a very lonely, disheartening position and at times
it is hard not to be bitter, sarcastic and vindictive. In at least two
other cases I can think of I eventually won my case but it sometimes
took years. Whether or not Dembski will be successful in getting his
ideas accepted is very unclear. Alienating people who might have been at
least somewhat inclined to read/consider his ideas also seems unwise.

Dave W

PS For an example of a negative result, I lead a feasibility study for
a new incremental compiler and associated development system that was at
the boundary between a research vrs development project. We also had
access to a sort of working, partial prototype that a couple of people
had been looking at for six months to a year. At least one IBM Fellow
who recently won the Turing award, endorsed the ideas. In the end the
study recommended we proceed and the project went ahead although there
were a minority of nay sayers. Unfortunately the project was not the
great success we hoped as it did not meet our primary goal of a
reduction in application build time, although it did meet quite a few of
the lessor goals and was shipped but then significantly improved in
later releases.

To unsubscribe, send a message to majordomo@calvin.edu with
"unsubscribe asa" (no quotes) as the body of the message.
Received on Wed Apr 25 16:35:13 2007