Id hate to be remembered for a dumb prediction that I didnt even make, but thats the posthumous predicament of Charles Duell. Commissioner of the U.S. Patent and Trademark Office in 1899, Duell never actually said that "everything that can be invented has been invented," notwithstanding 10,000 Web pages claiming that he did.
Id planned to begin this column by quoting that widely cited misprediction. I wanted to talk about areas of IT that are often said to be near their limits but where likely breakthroughs could disrupt that conventional wisdom.
I quickly discovered, though, that Duells dictum was a myth. An archivists study debunked the story in 1940, suggesting that it might have stemmed from an 1843 statement by then-Commissioner of Patents Henry Ellsworth. "The advancement of the arts, from year to year, taxes our credulity," Ellsworth said in his report that year to Congress, "and seems to presage the arrival of that period when human improvement must end."
You see what happens when you try to use ironic hyperbole? People take you literally. Ellsworths report concluded with projections of where patent activity would increase. He did not expect inventors to hit the wall; rather, he invited his readers to question whether exponential curves of improvement must always level off into the classic sigmoid curves of saturation.
More than 160 years later, were still wondering. In some cases, people warn that things are almost as good as they can get. In others, things are said to be so good that theres no motive to make them much better. In every case, I doubt these gloomy or complacent forecasts.
Three examples come to mind. The growing speed and shrinking size described by Moores Law are often said to be approaching dead ends. Crypto algorithms, the foundation of e-business, are said to be as good as theyll ever need to be. Software development rounds out my list of disciplines where progress seems in jeopardy because past trends have been so disappointing.
People challenge the demand for processor improvements. A faster machine, its said, merely waits more quickly. This looks like a temporary imbalance between unrecognized wants and emerging abilities—and the software writers are catching up.
For example, the latest Nikon digital cameras offer "face priority" focusing, using algorithms from the biometrics vendor Identix to spot the faces in a frame and make them sharp. Who would have thought to ask for this until someone offered it? But it will demand more CPU than cameras ever "needed" before.
As for any imminent repeal of Moores Law, silicon transistors arent the only switches in town. The crossbar latch technology disclosed this month by Hewlett-Packard extends the size of switching devices down to single-digit nanometers, compared with the tens of nanometers where current transistors play. Well have the CPU cycles that we need.
Current crypto algorithms, meanwhile, are only as strong as the presumed brick wall that blocks rapid factorization of the products of large prime numbers. Crypto experts concur that this is hard enough to make 1,024-bit keys sufficient—but, when pressed, theyll admit that 2,048 bits are better for long-term protection. Adi Shamir, the "S" in the RSA algorithm, estimated two years ago that a $15 million machine could crack a 1K-bit key in a year.
Its not far-fetched, moreover, to project that quantum computers will crack contemporary crypto before they do anything else. The problem is well-defined and well-suited to the emerging "qubit" technology. I therefore wouldnt want to declare the end of cryptos era of invention.
That brings us to my third example mentioned. Using faster processors to do interesting things securely is the task of software developers—whose productivity, unfortunately, has grown by only a few percent per year since the 1980s. Improved interfaces, team programming and enhanced reusability are clear pathways to improvement, so, again, I wont admit that theres nothing left to invent.
Id be happier, though, if I could say that without the backdrop of recent software project abandonments—which suggest that Henry Ellsworths "end of human improvement" might someday turn from rhetoric to reality.
Technology Editor Peter Coffee can be reached at firstname.lastname@example.org.