IP is Unique --- human involvement solves AND the legal system is functionally incapable of spillover.

Christopher Markou, “From Rule of Law to Legal Singularity,” Is Law Computable? https://www.academia.edu/44516947/From_Rule_of_Law_to_Legal_Singularity

On the vexed question of whether to attribute legal personality to AI systems, Abbot and Sarch warn against the danger of ‘rights creep’, as has happened already with the attribution, in a number of jurisdictions, of constitutionally-protected human rights to corporations. One solution, they suggest, would be to designate a ‘responsible person, in practice likely to be a corporation, as the ultimate guarantor of an AI. They also consider the costs and benefits of requiring mandatory registration and reporting for certain AI systems.

They conclude that notwithstanding ‘the growing possibility of Hard AI Crime’, the ‘radical tool of punishing AI’ would be an overreaction. Preferable would be ‘modest expansions to criminal law, including, most importantly, new negligence crimes centered around the improper design, operation, and testing of AI applications as well as possible criminal penalties for designated parties who fail to discharge statutory duties’.

8.8 Lyria Bennett Moses --- Not a Single Singularity

Lyria Bennett Moses chapter, ‘Not a Single Singularity’, envisages the development of AI in law occurring along three dimensions: an ‘x axis’ describing what is currently possible at any given time, a ‘y axis’ referring to what the technology is capable of becoming, and a ‘z axis’ along which decisions about the legitimacy of AI will be made. She argues that rather than there being a ‘straight path’ towards a legal singularity, there are likely to be periods of alternating progress and stasis along each of the three axes, and that the final outcome is unlikely to be a fully computable legal system.

In the context of legal AI, human and machine capabilities are more likely to be complements than substitutes. Machines can currently simulate certain human tasks, and if the focus is on observed functions they may be understood as simulating certain forms of rational behaviour. However, the current state of AI systems is such that they are very far from achieving anything approaching general intelligence. When particular techniques are closely scrutinised, their limitations with respect to law come into sharp relief. Thus attempts to make legislation machine-readable through the use of expert systems techniques have shown that it is most straightforward to code those statutory texts which embed an element of calculation or which rely on relatively straightforward decision rules. The danger in seeking to translate legislation into code is that it will change the way in which laws themselves are drafted. Expert systems cannot code for openended concepts such as ‘reasonableness’, but general clauses of this kind which, by virtue of their incompleteness, permit legal adaptation.

Similar problems arise when natural language processing and machine learning are used for case prediction. The current state of these techniques is inadequate given their tendency to replicate racial and other biases, as the case of the COMPAS software has highlighted. But even these capabilities can be enhanced in future, there are problems of legitimacy associated with their use. In the context of legal decision making, judgment is not a purely predictive process; it is the outcome of deliberation and contestation.

She concludes that an uneven development of AI in law is likely, not just because of limits to what be achieved technically, but by virtue of unavoidable concerns over the legitimacy of automated decision making. Technological advances ‘will lead to both progress and error, sometimes expanding what is available, what is possible and what is appropriate and legitimate in ways that are both evolutionary and revolutionary. But there are many thresholds to cross, and it is hard to imagine a system that would render law fully computable without changing the nature of law itself.’

8.9 Dilan Thampapilai --- The Law of Contested Concepts

Dilan Thampapilai’s chapter, ‘The Law of Contested Concepts? Reflections on Copyright Law and the Legal and Technological Singularities’, argues that while the idea of the legal singularity might seem convincing enough at a high level of abstraction, it runs up against serious objections in context of an applied field of law such as intellectual property (IP). IP law reflects a complex bargain around notions of property, human rights, free expression, technological development, and education, among other things. Copyright can only functions as an imperfect property system in which the courts have discretion to balance competing interests and values on a case by case basis.