IP doesn’t motivate innovation.

Reto Hilty, Jörg Hoffmann, and Stefan Scheuerer, “Intellectual Property Justification for Artificial Intelligence,” Artificial Intelligence and Intellectual Property, 04-19-2021, https://global.oup.com/academic/product/artificial-intelligence-and-intellectual-property-9780198870944?cc=cr&lang=en&

3.2.1.1 General incentive theory

If the incentive theory is generally not convincing, it loses even more relevancy in the AI context. The computer itself, ie, the AI tool, is not conscious of or responsive to incentives.68 Regarding the various human actors involved—whether as a programmer, a potentially creative or innovative user of an AI tool, or a distributor of AI-generated output69—a certain influence of legal incentives might be conceivable.70 But it is in no way apparent that there is a lack of motivation to be active in these areas.71 Rather, on a general level one can presently observe that AI innovation appears to be thriving. The markets are highly dynamic. Market failure is not apparent.

It has been claimed that the (human) programmers need to obtain rights to the results generated by AI, because programming would not be undertaken if third parties could immediately free ride.72 However, no evidence supports this assumption, and in any case it does not substantively go beyond the mere reiteration of theoretical, abstract lines of IP reasoning.73 A counter view argues that the programmer’s incentive to develop AI applications is sufficiently taken into account ‘higher up the chain’74 by IP protection for the elements of the specific AI tool75 or rather the sales or licensing revenues pertaining thereto.76 This line of argument, however, constitutes the economic mirror image, so to speak, of the deontological ‘double reward’ thought outlined above. Thus, it suffers from the same misguided assumption that IP protection for the tool as such is given or justified in the first place. In any case, all perspectives ultimately remain speculative for lack of empirical grounds.

With regard to humans using AI tools with a view to obtaining a superficially creative or potentially innovative output, then, ‘the mere action of pressing a button’77 (which is the ultimate technological promise of AI) is certainly not in need of external motivation. As far as subsequent distributors are concerned, a variety of business models to make profitable use of public domain subject matter are imaginable—reaping these possibilities is up to the commercial creativity of the market players.

At the same time, considerable AI-related patenting activity can be observed.78 However, there is no evidence concerning whether or to what extent such patents are actually obtained because they are associated with incentivizing functions. Rather it can be assumed that there are varieties of market-driven reasons (in particular first-mover advantages79 and the almost infinite advantages AI-enabled knowledge gathering possesses) that motivate undertakings to engage in AI markets. Consequently, they may be patenting for a variety of differing reasons, in particular for corporate strategy purposes.

In addition, there is an observable shift towards self-regulation tendencies. The more market players voluntarily share their intangible assets—whatever their actual motives may be—the less their actions can be incentivized by the prospect of obtaining or making use of state-granted exclusive rights. As shown above, in essence AI works on the basis of data and algorithms. In both fields, self-regulatory sharing schemes play a major role.

As regards software algorithms in particular, the ‘Open Source’ movement has long been a self-regulatory deviation from proprietary state IP law.80 Its ‘openness and sharing culture’ appears to particularly thrive also in the AI context: Large players like Google share and publish their new developments for ‘free’ (ie, at least without an explicit counter-performance).81 Optimization algorithms are to a large extent available in publicly accessible ‘libraries’ such as Google’s ‘TensorFlow’.82 Individual programmers may often be intrinsically motivated to share by reputational prospects.83

Also, with regard to data, pools and sharing platforms have gained considerable importance, and their practical relevance is expected to keep on growing.84 These, too, are operating under paradigms of self-regulation,85 especially with regard to access issues. Private ordering86 thus plays a major role in designing innovation enabling frameworks for AI markets. Of course, there is a highly complicated interaction between the existing IP regime and the self-regulatory schemes, partly revolving around, building on, or deviating from them. Nevertheless, in their specific appearances outlined above, self-regulation tendencies to a certain extent call into question IP’s exclusionary property logic and ultimately its justification in the realm of AI.

A point that is at least indirectly related to the incentive theory relates to the role of IP as a potential signal of firms’ innovativeness in attracting venture capital. A correlation between a firm’s innovation capacity and the number of patents it holds has traditionally been seen. However, there are new forms of investment chances, which relativizes the relevancy of patents above all for startups. Stateoffered research funds also allow direct market intervention to promote innovative or creative output. Direct research subsidizing can also facilitate knowledge transfer from science to practice.87