close icon-linkedin icon-twitter icon-facebook icon-mail icon-google-plus icon-search icon-phone icon-instagram
What are you looking for?
AI Chess
Blog 17 Jun 2020

Can artificial intelligence invent something?

Written by Rolf Suurmond

Blog series: Part 2

When clients ask me who they can name an inventor, I use to say, ‘anyone who made an intellectual contribution to the invention’. It never occurred to me to designate any tool that was used to create the invention (say, a calculator or a computer to run a simulation model) as a co-inventor.

However, if a machine has learnt by itself how to play chess, and has done it so successfully that no human being can defeat it? This is the case for current chess computers. Is the computer ‘the world’s best chess player’? And if a computer can be a ‘chess player’, why couldn’t a computer be an ‘inventor’, if it performs all the acts that are involved in ‘making an invention’? But wait, the chess computer was programmed by a human being with a detailed learning algorithm to learn the finesses of chess. The computer just blindly executed the learning algorithm that was programmed into it.

The same is true for making inventions; a mathematician designs a neural network that is executed by a computer, to program the computer to do a particular analysis task using a lot of data. I think we have to be very cautious before we accept that an artificial intelligence is the inventor. For example, is it sufficient if the artificial intelligence outputs a design or a chemical formula? Or should it be a requirement that the inventor/artificial intelligence is able to discuss about the invention on a conceptual level using natural language? The only certainty we have about this, is that the patent laws do not provide any answers to such questions.

Turning to current developments in patent practice, this topic is actually being explored in a practical case. It is not just science fiction. An applicant named an artificial intelligence called ‘DABUS’ as the inventor in a patent application. The European Patent Office took the point of view that they were not supposed to dispute that ‘DABUS’ actually made the invention, because they are to accept the information provided by the applicant. Nevertheless they refused the patent on formal grounds, because the Patent Convention suggests that the inventor must be a human being. The case is going into appeal.

Next week, I’ll dig a bit deeper in the legal question whether the patent laws would allow the mention of a machine as an inventor in a patent application.