Fordham 27 (Report 7): AI

The AI panel 

Having finished lunch on the first day, the Fordham IP conference attendees were ready for some more topic-specific sessions. The AmeriKat hands over to one of the Fordham guest Kats, Alexander de Leeuw (Brinkhof) for the report on the AI session:

Over to Alexander:

"The session on artificial intelligence, moderated by Laura Sheridan (Google), appeared to have drawn particular attention given that people were standing in line all the way outside the conference room to get a glimpse of what the experts had to say.

The first speaker to take the stage was Massimo Sterpi (Gianni, Origoni, Grippo Cappelli & Partners), who talked about neurography and “the rise of the neural artist”. After diving into the history and technicalities of AI, Massimo explained that there are many humans involved in the process of creating output through AI, such as those who create the structure, those who provide the content, those who train the adversarial networks, and those who utilize the output (see Kat post by Alexander here).  A central theme of Massimo’s presentation was logically who, then, should be able to obtain some form of IP protection?  As correctly indicted, all disputes regarding AI so far were not about the presence or absence of IP, but about entitlement to IP. The cases have been about humans claiming recognition, not the more fundamental question of whether there should/could be IP protection for AI.

Massimo proposed that whoever had an intervening role in the process of creating output through AI should be involved in sharing the profits. His presentation concluded with three options by which this could be achieved in the future:

• moving from a romantic model of authorship into a more curatorial model;
• moving to a sui generis right where the focus is on the effort of the creation rather than on the author’s expression; or
• moving to database protection for the database underlying the AI output.

Second to speak on this topic was Carlo Scollo Lavizzari (Lenz Caemmerer), who spoke about “how machines learn and what it means for authors, publishers and media businesses”. One of the prominent questions raised by Carlo is why AI is currently such a hot topic. Carlo addressed this by stating that what has changed is the powerful combination of computing power and the big data environment. One of the more interesting propositions of this presentation was about how the market might be changing. AI needs to be trained with data. In fact, without data, software cannot be considered AI; the link between the two is data. According to Carlo, “if training machines with data becomes such a big market, that should become part of the normal exploitation of a work. Machines may review papers more often than doctors. The market will effectively be the machines. Perhaps we should be protecting the data used in machine learning”.

Katherine Stephens (Bird & Bird) then talked about the “patentability of artificial intelligence and machine learning”, specifically focusing on the recently updated EPO Guidelines for Examination. According to Katherine, the EPO’s new Guidelines are not a green light for patenting AI, but they are a first step in setting out the rules for a proper balancing exercise. An interesting issue raised in her presentation was whether inventive step and sufficiency thresholds can be expected to change with the rise of AI, assuming that the skilled person should be presumed to have access to AI systems. “Will inventive step be raised so high that nothing will be considered inventive in the eyes of the law, even if it was inventive or surprising to human?”

During the debate at the end of the session the more fundamental policy question was addressed of whether and why society would want to provide IP protection for AI. Massimo interestingly suggested that “a machine does not need to be incentivized. It simply creates. Does it need copyright protection?” Mihály Ficsor (Hungarian Copyright Council) suggested that it would perhaps be more appropriate to provide a shorter term of protection and highlighted repeatedly that this discussion is an old discussion that was starting with the creation of software. Katherine Stephens questioned what the economic benefit is of affording IP protection to AI and Shlomit Yanisky-Ravid (ONO Academic College) noted that “providing copyright to non-human works brings with it a lot of questions”.

Massimo finished the debate with the remark that “ultimately, whatever outcome AI generates is still within the space of possibilities the creator of the AI defines”. It remains to be seen whether AI will evolve to develop its own artistic personality, will at some point become largely independent of human interaction, and whether and how this requires changes to IP legislation."