A few days ago, April 24th, was the 20th anniversary of the launching of the Hubble Space Telescope.  arsTechnica celebrated with a nice article here, from which I will shamelessly lift this famous picture:

"My god, it's full of galaxies!"

This patch of the sky is so tiny that it only contains a few stars from our galaxy, but is packed with galaxies without end.  There’s ten billion light years in a glance.   The total exposure time is a million seconds, done in 800 exposures of about 20 minutes each.

After 20 years of service, the Hubble is the most productive single instrument in the history of science.  That is, more peer-reviewed papers have been written based on its data than any other instrument (~9000 at last count) , and those papers have a significantly higher impact factor (they get cited more often) than typical papers.   Almost 3000 scientists have contributed to those papers, a significant percentage of all the astronomers in the world.   Its biggest discovery is that the Hubble Constant is not – galaxies do not recede as a linear function of distance, but at a more than linear rate.  This is the so-called “dark energy” that no one understands, but are aching to find out.

The Hubble has also been a public relations triumph for NASA and America.  PR is the main justification for NASA’s huge budget, and Hubble has paid off in that regard in a way that the Shuttle and the International Space Station have not.   The Shuttle is ugly and the ISS is purpose-less, but Hubble keeps delivering these gorgeous images.  Yes, its mirror was mis-ground by Perkin-Elmer, but heroic effort by engineers and astronauts saved the mission.  The latest servicing mission was web-cast live, and was fascinating.  There’s now even an IMAX movie about it.

Yet for all that, Hubble is being overtaken by ground-based and balloon-based scopes.   Adaptive optics gives ground-based scopes better angular resolution than Hubble, and with much larger mirrors.  (This technique, by the way, has been the only major result of the 50 years of the Pentagon’s advanced research group, the JASONs.  This is a group of leading civilian scientists that meet each summer to discuss issues of note.  You would think that they would meet in a bunker in Area 51 where the aliens can’t get them, but they like to meet in San Diego so their kids can play  on the beach.)  Balloon-based scopes like the Sunrise and BLAST can see into the ultraviolet and infrared just like Hubble and for vastly less cost.  Right now they can only fly for a few days at a time, but once they start using solar-powered robot planes (E.g. HELIOS), they’ll be able to orbit Antarctica for the six months of daylight.  Building such planes are expensive, and they might cost as much as TEN MILLION DOLLARS, as Dr. Evil would say, but that’s chump change by space standards.  This paper compares the number of papers in Nature that came out each of the major scopes in the period 1991 to 1998, and finds that Hubble leads the pack, but not by much.

You see this over and over again in engineering.  Someone tries something really out of the mainstream, in the hope of exploiting some big natural advantage, and they get ultimately crushed because so many people are working on the standard approach that they can’t compete.  RISC vs the x86 architecture is a good example.   Reduced Instruction Set Computing looked like a naturally better way to design faster computers, but it ultimately failed.  Intel and AMD could put so much talent into speeding up the hideous x86 instruction set that RISC ultimately had little performance advantage.  Worse, it had a huge disadvantage in the lack of software.   There was a big return-on-investment of speeding up the computer that was on everyone’s desk, and not a great ROI in forcing everyone to re-do their software for only a 2X speedup.

Putting a big telescope in space was a vastly expensive project, in the $5 billion to $10 billion range depending on how you price the Shuttle service missions.  Yes we love the images it has returned, and admire the craft of the project itself, but the science itself could probably have been done for a lot less.    A lot of ingenuity applied to a lot of cheap scopes will give a better scientific return than a lot of money spent on a big showcase project like Hubble.