Daniel Markham takes me to task for being one of those “software patents are destroying the world!” types:
Imagine for a minute that I just got off a time machine from the year 5600. I know how to make truly intelligent machines, so I sit down and write a patent on how to make computer intelligence. Now at the heart of my patent will be arrays, indexes, memory cores–all of the usual computer stuff. It’s all just ones and zeros, folks. But obviously my patent has tremendous value to society. This is a silly example, but since it IS possible to make an example where software patents make sense, the question isn’t whether they are useful or not, the question is how to tell the difference. That’s a big point that a lot of folks miss. Get rid of the bath water, keep the baby.
This is a silly example for a number of reasons, and not just the obvious ones. In the first place, it’s unlikely that somebody’s going to sit down at his computer and come up with a single breakthrough that makes computers instantly intelligent. More likely, there will be a long series of incremental improvements. Each advancement will give its creator a short-term advantage in the marketplace before another firm comes up with another incremental improvement that puts it ahead. This process of incremental improvement and imitation is the way the software industry has worked for decades.
In that circumstance, it would be absurd to give the first guy to write a program that passes the Turing test a 17-year monopoly on all AI. Had he not done it, chances are a dozen other people would have within a few short years. He should only be entitled to a monopoly on his particular incremental improvements–and that’s precisely what he gets under copyright law.
Moreover, if you were going to give the guy a patent, what would its scope be? Such a patent would probably require dozens of pages of description to explain how it all works. If somebody else writes a different AI program that uses some of the same techniques and some different techniques, a judge would have to read through hundreds of pages of source code to decide if the new program was similar enough to the patent to qualify as infringing. How could there possibly be clear standards in a situation like that?
Most importantly, you don’t need software patents to ensure the inventor of the first AI program gets rich. Small, nimble startups are constantly getting rich using innovative software ideas without the aid of software patents. The marketplace gives tremendous advantages to the first company to introduce a particular type of product or service. Yes, at some point big software companies will introduce imitations, but they only start developing such imitations after the startup has gained considerable momentum. Moreover, the big companies will always be behind: they can imitate the current version of the startup’s project, but by the time they get that out the door the startup is likely getting ready to release another version.
Finally, Markham says “it’s possible to make an example where software patents make sense.” Yet all he can come up with is a goofy hypothetical example. If software patents are a good idea, then surely Markham (or other software patent defenders) can come up with at least one example of a good software patent in the real world!