Saturday, March 26, 2016

Can a Computer Get a Patent?

Artificial intelligence (AI) has been in the news this week. Yesterday's New York Times story about AI beating a human at Go earlier this month declared: The Race Is On to Control Artificial Intelligence, and Tech’s Future. That race may be a slow one: Microsoft's AI Twitter bot Tay was quickly hijacked by trolls, and the Japanese AI heralded for writing an almost-awarding winning novel actually did little of the creative work. But at some point, it seems plausible that AI will make independent contributions to creative and inventive processes—which, for IP lawyers, raises the question: who owns the IP?

Most takes on this problem have been primarily on the copyright side; see, e.g., pieces by Pam Samuelson (1986), Ralph Clifford (1997), Annemarie Bridy (2012), and James Grimmelmann (forthcoming). A recently posted article by Ryan Abbott, I Think, Therefore I Invent: Creative Computers and the Future of Patent Law, tackles the problem primarily from the patent side. Abbott argues that the USPTO has already (unknowingly) granted patents for inventions by computers (such as the Creativity Machine and the Invention Machine), and that such patents should belong to the owner of the inventive computer, rather than the software developer or user.

I'm not sure what the most efficient default is, or whether computers actually are making contributions such that it is sensible to describe the computer as an inventor independent of its programmers. But even if it is not an immediate problem, it would seem prudent for innovators working with AI to follow the same practices as those working with other humans: make sure the relevant contracts are clear on who owns any resulting IP.