To patent or not to patent discoveries from artificial intelligence, that is the question.
Amanda D. Francoeur, firstname.lastname@example.org
By using the scientific method – careful observation, formation of hypotheses, experimentation, interpretation and reproduction of the results – a researcher has given us a better understanding of the genomics of the yeast Saccharomyces cerevisiae, including new information about gene sequencing and the encoding of orphan enzymes. But what is particularly new is that the researcher was not a human but a robot named Adam.
Adam was created by a team led by Ross D. King, a professor in the computer science department at Aberystwyth University in the UK. Working autonomously, the robot formulated and tested 20 of its own hypotheses in connection with 13 orphan enzymes and observed 6,657,024 optical density measurements. It also designed and initiated more than 1000 new yeast strains and defined growth medium experiments lasting up to five days. Adam’s most important discovery was that three genes (YER152C, YJL060W and YGL202W) encode the enzyme 2-aminoadipate:2-oxoglutarate aminotransferase (2A2OA), which catalyzes a reaction in lysine biosynthetic pathways of fungi. To confirm Adam’s conclusions, King purified the protein products of the genes and applied them to in vitro enzyme assays.
Knowledge of gene-encoding enzymes remains incomplete, but Adam’s use of systematic bioinformatics and phenotypic analysis marks a turning point for researchers, as they will be able to explore comparative genomics in detail.
Laying down the law
In a letter published in the May 22, 2009, edition of Science, registered patent attorney Robert W. Stevenson and colleagues from Caesar, Rivise, Bernstein, Cohen & Pokotilow Ltd. in Philadelphia questioned whether ideas generated by a machine could be patented. They cited 35 U.S.C. Section 102(f), which states that a “person shall be entitled to patent.” The code also states that an inventor cannot obtain a patent if “he did not himself invent the subject matter sought to be patented.”
It seems, therefore, that under US law, only discovery by a human being can be given full protection under the law; an idea from a different informant – e.g., a machine – may not obtain valid patent protection.
In this situation, one also could apply Section 101, which says that “whoever invents … may obtain a patent,” not “whatever,” further specifying that patents are issued to human beings.
Stevenson brings up another issue related to patentable subject matter: the requirement of nonobviousness. “If a machine can come up with a theory based on how it’s programmed, would the court consider the invention obvious?” An obvious invention is an idea that would be evident to a person with ordinary skill in the subject matter at hand, much as King’s knowledge of gene sequencing enabled him to create a robot that can analyze genes.
There is a certain amount of “care that should be taken when applying these technologies,” Stevenson said. At least by supplying your own theories to the machine and allowing it to confirm the development later, “the idea is still yours, and there shouldn’t be any problem with the patent law that would exclude you from obtaining a valid patent,” he added.
It is hard to speculate whether the US Patent and Trademark Office (USPTO) would ever grant patentability to discoveries such as Adam’s or whether the inventions would become freely available without patent protection. Nevertheless, in an age where robotics technology is advancing at an exceedingly fast rate and computers are increasingly becoming a staple in scientific experimentation, it is quite probable that machines akin to King’s robot could redefine the term “inventor” or even rewrite the book on patent law.