Multidimensionality of Legal Singularity: Parametric Analysis and the Autonomous Levels of AI Legal Reasoning Dr. Lance B. Eliot Chief AI Scientist, Techbruim; Fellow, CodeX: Stanford Center for Legal Informatics Stanford, California, USA Abstract important parametric analysis of the Legal Singularity. Legal scholars have in the last several years embarked The final section, Section 5, covers additional upon an ongoing discussion and debate over a considerations and future research. potential Legal Singularity that might someday occur, involving a variant or law-domain offshoot leveraged This paper then consists of these five sections: from the Artificial Intelligence (AI) realm amid its • Section 1: Background of The Singularity many decades of deliberations about an overarching • Section 2: Legal Singularity and generalized technological singularity (referred to classically as The Singularity). This paper examines • Section 3: Autonomous Levels of AI Legal the postulated Legal Singularity and proffers that such Reasoning AI and Law cogitations can be enriched by these three • Section 4: Legal Singularity Multidimensionality, facets addressed herein: (1) dovetail additionally Alignment with LoA AILR salient considerations of The Singularity into the Legal • Section 5: Additional Considerations and Future Singularity, (2) make use of an in-depth and Research innovative multidimensional parametric analysis of the Legal Singularity as posited in this paper, and (3) align Since the word “singularity” is used in at least two and unify the Legal Singularity with the Levels of contexts within this paper, one context being an Autonomy (LoA) associated with AI Legal Reasoning overarching or grandiose kind of singularity, typically (AILR) as propounded in this paper. referred to as The Singularity, and the other being a Keywords: AI, artificial intelligence, autonomy, singularity specific to the field of law, known as the autonomous levels, legal reasoning, law, lawyers, Legal Singularity, the convention in this paper will be practice of law, The Singularity, Legal Singularity that whenever referring to the Legal Singularity this will be done by stating “Legal Singularity” or by the abbreviation of “LS,” while the larger The Singularity 1 Background of The Singularity will be referred to as the “singularity” or “AI singularity” or “Technological singularity,” and when In section 1 of the paper, the topic of The Singularity desiring to especially emphasize such a reference it is introduced and addressed. Doing so establishes the will be stated as The Singularity (such an emphasize is groundwork for section 2, covering a form of done solely for drawing attention to the matter and not singularity that has come to be known as the Legal due to suggesting any differences of meaning or Singularity (LS), considered to be an offshoot or a connotation). domain-specific variant of the overarching The Singularity. Section 3 indicates the Levels of 1.1 Understanding The Singularity Autonomy (LoA) of AI Legal Reasoning (AILR), which will be instrumental in the discussions A longstanding discussion and debate in the field of undertaken in Section 4. Section 4 then provides an in- Artificial Intelligence entails a controversy referred to depth analysis of the Legal Singularity as it relates to as The Singularity [9] [14] [28] [39]. Sometimes also the LoA of AI Legal Reasoning and lays out an 1 coined as the AI Singularity or the Technological Singularity, the concept underlying the matter is Here is Good’s [36] definition associated with the relatively ill-defined and has substantively varied in capabilities envisaged: “Let an ultra-intelligent details of its meaning and substance over the now machine be defined as a machine that can far surpass many years of its postulation (dating back to the all the intellectual activities of any man however 1950s). clever. Since the design of machines is one of these intellectual activities, an ultra-intelligent machine Often first-traced to commentary by the famous could design even better machines; there would then mathematician and pioneering computer scientist John unquestionably be an intelligence explosion, and the von Neumann, here is what researcher Ulman [58] in intelligence of man would be left far behind. Thus, the 1958 indicated had occurred in a conversation with first ultra-intelligent machine is the last invention that Von Neumann: “One conversation centered on the man need ever make.” ever-accelerating progress of technology and changes in the mode of human life, which gives the appearance In this initial elucidation of the topic by Good, a key of approaching some essential singularity in the facet that has become inextricably woven into the history of the race beyond which human affairs, as we singularity rubric is the speculative idea of an know them, could not continue.” intelligence explosion. In short, if mankind can craft an AI system to some threshold of intelligence, it is Essentially, the sentiment at the time was that presumed that the AI could then further progress, computers might eventually be able to achieve human essentially on its own accord, by using its core base of intelligence, potentially even eclipsing human intelligence to further produce more intelligence. No intelligence, and the result could be problematic for one as yet knows what this minimum threshold might humanity. Of course, similar exhortations have been be, and nor is there any viable means to anticipate how replete in science fiction, though typically proffered by far the presumed intelligence explosion might proceed imaginative writers with unsupported visions rather in terms of the upper limits of some unknown super- than by bona fide scientists that are making such intelligence, raising the perennial question of how high speculations based on their assessment of the is up, as it were. underlying technology and attempting to anticipate future outcomes. That’s not to suggest that those Pursuing for the moment the somewhat tangential but scientists will necessarily be on par with predicting the relevant question concerning the notion of an future, and there are many documented instances of intelligence explosion, let’s consider the ramifications scientists that were wildly off-the-target and baseless of such a phenomenon, if indeed possible (no one in their prophesizing. In short, expertise in a subject knows whether it is or not). Similar to questions that matter is a worthwhile basis for providing meaningful arose during the creation of the atom bomb, whereby predictions, nonetheless, that expertise can still be scientists were somberly worried that the ignition and misguided or mistaken as to what the future might exploding of an atomic bomb might somehow catch hold. hold and violently and rapidly spread across the globe in an unheralded conflagration, some assert that the In 1965, Oxford researcher Irving John Good [36] same might happen in the case of an intelligence published a cornerstone research paper that extended explosion. To wit, the intelligence being produced the singularity notion and tied the topic to the might magnify and expand, for which the result could emergence of computers that might be considered as be to have humans seem like mere ants in intelligence ultra-intelligent, commonly today referred to as aiming versus the super-intelligence spawned by AI. Again, to be super-intelligent or super-human in capability. some view this outcome as disastrous for humanity, Rather than emphasizing the dangers associated with possibly being enslaved by a super-intelligent AI, mankind developing an ultra-intelligent machine, while others believe that mankind might be saved due Good [36] urged that the survival of humanity to an artificial super-intelligence that could solve the depended on being able to craft such a system and gravest problems confronting the survivability of indeed ought to be done as soon as possible: “The humans. survival of man depends on the early construction of an ultra-intelligent machine.” 2 In his research, Good attempted to outline some of the Arguments ensue as to which of those is the overall features or elements that seemed at the time to “singularity” and also whether they must be combined potentially be required to achieve an AI super- or co-existent to count as the singularity occurring. intelligence. For example, he debated those theories of Perhaps “a” or arrival at human intelligence is the period concerning a tremendous amount of mandatory for getting to “b,” though others contend parallelism that would be needed, a facet of modern- that it is possible that an AI decidedly less-than-human day computers that were not especially viable when intelligence levels might percolate via “c” per an Good [36] wrote his paper in the 1960s: “It cannot be intelligence explosion and then nearly instantaneously regarded as entirely certain that an ultra-intelligent exceed “a” and arrive at “b,” thus never especially machine would need to be ultraparallel since the settling down at the mere capacity of human number of binary operations per second performed by intelligence. Others argue that there is not anything the brain might be far greater than is necessary for a feasibly beyond human intellectual capacities, computer made of reliable components. Neurons are regardless of how adept the AI might be, and as such not fully reliable; for example, they do not all last a the achievement of human intelligence is the capstone lifetime; yet the brain is extremely efficient. This limit. In that case, the singularity would be solely efficiency must depend partly on ‘redundancy’ in the about the “a” and not take into account the “b” and “c” sense in which the term is used in information theory. postulates which are deemed as impossible and a false A machine made of reliable components would have aspiration. an advantage, and it seems just possible that ultraparallel working will not be essential. But there is Additionally, some assert that a super-intelligence a great waste in having only a small proportion of the might be reached without any need for and indeed no components of a machine active at any one time.” occurrence at all of an intelligence explosion.