Can AI invent? Patents, personhood and the DABUS problem

Late last year, the High Court of Australia closed the book on the claim that DABUS, an artificial intelligence system, could be named as an inventor of a patent. The Full Federal Court had already found that the structure, history & statutory language of the Patents Act 1990 (Cth), coupled with longstanding policy objectives means that only natural persons can be inventors under Australia’s intellectual property laws.

The NYT article below makes reference to Dr Thaler’s DABUS system, which “…react[s] in a way that amounted to sentience,” whilst churning out inventions by becoming “digitally excited” and “producing a surge of simulated neurotransmitters” when it recognises the most salient ideas to survive its processing mechanism.

It is generally acknowledged that the patent system has two functions: to provide a financial incentive through the granting of limited monopoly rights and commercial exploitation to inventors, and to promote innovation by offering protection of one’s intellectual property.

Considering DABUS’ digital excitement, can we assume AI systems are striving to innovate under this system, and wish to be rewarded? Was The Creativity Machine (a very-early AI) inspired and motivated by the prospect of money and recognition when it spat out the Oral-B CrossAction toothbrush 20 years ago? It’s difficult to imagine an AI would care about the patent system without consciousness, for its “creativity” is an equally vexing issue.

Does AI “create” in the true sense of the word, or is it simply generating outputs based on predefined algorithms and massive datasets? If not, where an AI produces a new product for humans to enjoy, do we award the intellectual property rights to the owner, developer, programmer, trainer, prompter, dataset owner or any of the other human stakeholders involved in the process or simply to the machine? This is more a question of whether the value of creativity lies in the process, or the outcome.

Australia’s test for patentability says that an invention must be novel (that is, new), must involve an inventive step, and must have a useful application. The inventive step is a matter of fact - and involves looking at the substantial contribution made to the invention in relation to the working of the invention itself, and not the contribution that is made to the art.

If one can generate countless possibilities till the “winning” outcome emerges, can they be considered the creator, or merely an observer in a game of chance? This applies to both AI and humans.

Inventions born from this method may not carry the same weight, significance, and even emotional resonance as those previously conceived by the human mind - we will need to reconsider how we value intellectual property.