Three Laws Blog: Law and Artificial Intelligence

Bilokon on Evolution-Boostrapped Simulation: Artificial or Human Intelligence: Which Came First? @bilokon @imperialcollege

Paul Bilokon, Imperial College Lndon, has published Evolution-Bootstrapped Simulation: Artificial or Human Intelligence: Which Came First? Here is the abstract.

Humans have created artificial intelligence (AI), not the other way around. This statement is deceptively obvious. In this note, we decided to challenge this statement as a small, lighthearted Gedankenexperiment. We ask a simple question: in a world driven by evolution by natural selection, would neural networks or humans be likely to evolve first? We compare the Solomonoff--Kolmogorov--Chaitin complexity of the two and find neural networks (even LLMs) to be significantly simpler than humans. Further, we claim that it is unnecessary for any complex human-made equipment to exist for there to be neural networks. Neural networks may have evolved as naturally occurring objects before humans did as a form of chemical reaction-based or enzyme-based DNA computation. Now that we know that neural networks can pass the Turing test and suspect that they may be capable of superintelligence, we ask whether the natural evolution of neural networks could lead from pure evolution by natural selection to what we call evolution-bootstrapped simulation. The evolution of neural networks does not involve irreducible complexity; would easily allow irreducible complexity to exist in the evolution-bootstrapped simulation; is a falsifiable scientific hypothesis; and is independent of / orthogonal to the issue of intelligent design.

Download the essay from SSRN at the link.

Sarel on Restraining ChatGPT @dr_roee @unihh

Roee Sarel, Institute of Law and Economics, University of Hamburg, is publishing Restraining ChatGPT in the UC Law SF Journal. Here is the abstract.

ChatGPT is a prominent example of how Artificial Intelligence (AI) has stormed into our lives. Within a matter of weeks, this new AI—which produces coherent and human-like textual answers to questions—has managed to become an object of both admiration and anxiety. Can we trust generative AI systems, such as ChatGPT, without regulatory oversight? Designing an effective legal framework for AI requires answering three main questions: (i) is there a market failure that requires legal intervention? (ii) should AI be governed through public regulation, tort liability, or a mixture of both? and (iii) should liability be based on strict liability or a fault-based regime such as negligence? The law and economics literature offers clear considerations for these choices, focusing on the incentives of injurers and victims to take precautions, engage in efficient activity levels, and acquire information. This Article is the first to comprehensively apply these considerations to ChatGPT as a leading test case. As the United States is lagging behind in its response to the AI revolution, I focus on the recent proposals in the European Union to restrain AI systems, which apply a risk-based approach and combine regulation and liability. The analysis reveals that this approach does not map neatly onto the relevant distinctions in law and economics, such as market failures, unilateral versus bilateral care, and known versus unknown risks. Hence, the existing proposals may lead to various incentive distortions and inefficiencies. The Article, therefore, calls upon regulators to place a stronger emphasis on law and economics concepts in their design of AI policy.

Download the article from SSRN at the link.

Katyal on Democracy and Distrust in an Era of Artificial Intelligence @sonia__katyal @americanacad @BerkeleyLaw

Sonia Katyal, University of California, Berkeley, School of Law, is publishing Democracy and Distrust in an Era of Artificial Intelligence in Daedalus, Journal of the American Academy of Arts & Sciences (2022). Here is the abstract.

Our legal system has historically operated under the general view that courts should defer to the legislature. There is one significant exception to this view: cases in which it appears that the political process has failed to recognize the rights or interests of minorities. This basic approach provides much of the foundational justifications for the role of judicial review in protecting minorities from discrimination by the legislature. Today, the rise of AI decision-making poses a similar challenge to democracy’s basic framework. As I argue in this essay, the rise of three trends–privatization, prediction, and automation in AI–have combined to pose similar risks to minorities. In this essay, I outline what a theory of judicial review would look like in an era of artificial intelligence, analyzing both the limitations and the possibilities of judicial review of AI. Here, I draw on cases in which AI decision-making has been challenged in courts, to show how concepts of due process and equal protection can be recuperated in a modern AI era, and even integrated into AI, to provide for better oversight and accountability.

Download the essay from SSRN at the link.

Stix on The Ghost of AI Governance Past, Present and Future: AI Governance in the EU @charlotte_stix

Charlotte Stix, Eindhoven University of Technology; University of Cambridge, Leverhulme Centre for the Future of Intelligence, has published The Ghost of AI Governance Past, Present and Future: AI Governance in the European Union. Here is the abstract.

The received wisdom is that artificial intelligence (AI) is a competition between the US and China. In this chapter, the author will examine how the European Union (EU) fits into that mix and what it can offer as a ‘third way’ to govern AI. The chapter presents this by exploring the past, present and future of AI governance in the EU. Section 1 serves to explore and evidence the EU’s coherent and comprehensive approach to AI governance. In short, the EU ensures and encourages ethical, trustworthy and reliable technological development. This will cover a range of key documents and policy tools that lead to the most crucial effort of the EU to date: to regulate AI. Section 2 maps the EU’s drive towards digital sovereignty through the lens of regulation and infrastructure. This covers topics such as the trustworthiness of AI systems, cloud, compute and foreign direct investment. In Section 3, the chapter concludes by offering several considerations to achieve good AI governance in the EU.

Download the article from SSRN at the link.

McCarl on The Limits of Law and AI @UCinLawReview @UCLALawReview

Ryan McCarl, University of California, Los Angeles, School of Law, is publishing The Limits of Law and AI in volume 90 of the University of Cincinnati Law Review (2022). Here is the abstract.

For thirty years, scholars in the field of law and artificial intelligence (AI) have explored the extent to which tasks performed by lawyers and judges can be assisted by computers. This article describes the medium-term outlook for AI technologies and explains the obstacles to making legal work computable. I argue that while AI-based software is likely to improve legal research and support human decisionmaking, it is unlikely to replace traditional legal work or otherwise transform the practice of law.

Download the article from SSRN at the link.

Colonna on AI in the Internet of Health Things @lianecolonna

Liane Colonna, Stockholm University Faculty of Law, has published Artificial Intelligence in the Internet of Health Things: Is the Solution to AI Privacy More AI? as Faculty of Law, Stockholm University Research Paper No. 90. Here is the abstract.

The emerging power of Artificial Intelligence (AI), driven by the exponential growth in computer processing and the digitization of things, has the capacity to bring unfathomable benefits to society. In particular, AI promises to reinvent modern healthcare through devices that can predict, comprehend, learn, and act in astonishing and novel ways. While AI has an enormous potential to produce societal benefits, it will not be a sustainable technology without developing solutions to safeguard privacy while processing ever-growing sets of sensitive data. This paper considers the tension that exists between privacy and AI and examines how AI and privacy can coexist, enjoying the advantages that each can bring. Rejecting the idea that AI means the end of privacy, and taking a technoprogressive stance, the paper seeks to explore how AI can be actively used to protect individual privacy. It contributes to the literature by reconfiguring AI not as a source of threats and challenges, but rather as a phenomenon that has the potential to empower individuals to protect their privacy. The first part of the paper sets forward a brief taxonomy of AI and clarifies its role in the Internet of Health Things (IoHT). It then addresses privacy concerns that arise in this context. Next, the paper shifts towards a discussion of Data Protection by Design, exploring how AI can be utilized to meet this standard and in turn preserve individual privacy and data protection rights in the IoHT. Finally, the paper presents a case study of how some are actively using AI to preserve privacy in the IoHT.

Download the article from SSRN at the link.

Houser on AI and the Struggle Between Good and Evil @kimberlyhouser @washburnlj

Kimberly Houser, University of North Texas, is publishing Artificial Intelligence and The Struggle Between Good and Evil in volume 60 of the Washburn Law Journal. Here is the abstract.

Numerous reports have described—in great detail—the real and potential harms of the widespread development and adoption of artificial intelligence by both government and private industry. However, artificial intelligence has also been shown to create faster, more accurate and more equitable outcomes than humans in many situations. This seeming contradiction has led to dichotomous thinking describing artificial intelligence as either good or evil. Artificial intelligence, like all technological developments, is a tool: one that can be used—intentionally and sometimes, unintentionally—for good or for harm. The gray area comes into play when there is good or neutral intent that leads to a harmful result. This Essay is designed to help policymakers, investors, scholars, and students understand the multifaceted nature of artificial intelligence and the key challenges it presents and to caution against creating laws that prohibit AI programs outright rather than addressing the fundamental need to develop AI responsibly.

Download the article from SSRN at the link.

Tschider on Preempting the Artificially Intelligent Machine @cybersimplesec

Charlotte Tschider, University of Nebraska, Lincoln, College of Law, has published Preempting the Artificially Intelligent Machine. Here is the abstract.

The medical device industry and new technology start-ups have dramatically increased investment in artificial intelligence (AI) applications, including diagnostic tools and AI-enabled devices. These technologies have been positioned as reducing climbing health costs while simultaneously improving health outcomes. Technologies like AI-enabled surgical robots, AI-enabled insulin pumps, and cancer detection applications hold tremendous promise, yet without appropriate oversight will likely pose major safety issues. While preventative safety measures may reduce risk to patients using these technologies, effective regulatory-tort regimes also permit recovery when preventative solutions are insufficient. The Food and Drug Administration (FDA), the administrative agency responsible overseeing the safety and efficacy of medical devices, has not effectively addressed AI system safety issues for its clearance processes. If the FDA cannot reasonably reduce the risk of injury for AI-enabled medical devices, injured patients should be able to rely on ex post recovery options, as in products liability cases. However, the Medical Device Amendments Act (MDA) of 1976 introduced an express preemption clause that the U.S. Supreme Court has interpreted to nearly foreclose liability claims, based almost completely on the comprehensiveness of FDA clearance review processes. At its inception, MDA preemption aimed to balance consumer interests in safe medical devices with efficient, consistent regulation to promote innovation and reduce costs. Although preemption remains an important mechanism for balancing injury risks with device availability, the introduction of AI software dramatically changes the risk profile for medical devices. Due to the inherent opacity and changeability of AI algorithms powering AI machines, it is nearly impossible to predict all potential safety hazards a faulty AI system might pose to patients. This article identifies key preemption issues for AI machines as they affect ex ante and ex post regulatory-tort allocation, including actual FDA review for parallel claims, bifurcation of software and device reviews, and dynamics of the technology itself that may enable plaintiffs to avoid preemption. This Author then recommends an alternative conception of the regulatory-tort allocation for AI machines that will create a more comprehensive and complementary safety and compensatory model.

Download the article from SSRN at the link.

Asay on Artificial Stupidity

Clark D. Asay, Brigham Young Law School, is publishing Artificial Stupidity in volume 61 of the William & Mary Law Review (2020). Here is the abstract.

Artificial intelligence is everywhere. And yet, the experts tell us, it is not yet actually anywhere. This is so because we are yet to achieve true artificial intelligence, or artificially intelligent systems that are capable of thinking for themselves and adapting to their circumstances. Instead, all the AI hype — and it is constant — concerns rather mundane forms of artificial intelligence, which are confined to performing specific, narrow tasks, and nothing more. The promise of true artificial intelligence thus remains elusive. Artificial stupidity reigns supreme. What are the best set of policies to achieve true artificial intelligence? Surprisingly, scholars have paid little attention to this question. Scholars have spent considerable time assessing a number of important legal questions relating to artificial intelligence, including privacy, bias, tort, and intellectual property issues. But little effort has been devoted to exploring what set of policies are best suited to helping artificial intelligence developers achieve greater levels of innovation. And examining such issues is not some niche exercise, since artificial intelligence has already or soon will affect every sector of society. Hence, the question goes to the heart of future technological innovation policy more broadly. This Article examines this question by exploring how well intellectual property rights promote innovation in artificial intelligence. I focus on intellectual property rights because these are often viewed as the most important piece of United States innovation policy. Overall, I find that intellectual property rights, particularly patents, are ill-suited to promote radical forms of artificial intelligence innovation. And even the intellectual property forms that are a better fit for artificial intelligence innovators, such as trade secrecy, come with problems of their own. In fact, the poor fit of patents in particular is likely to contribute to heavy industry consolidation in the AI field, and heavy consolidation in an industry is typically associated with lower levels of innovation than ideal. I conclude by arguing, however, that neither strengthening AI patents rights nor looking to other forms of law, such as antitrust, holds much promise in achieving true artificial intelligence. Instead, as with many earlier radical innovations, significant government backing, coupled with an engaged entrepreneurial sector, is at least one key to avoiding enduring artificial stupidity.

Download the article from SSRN at the link.

Barabas on Re-Imagining the Terms of "Ethical AI" in Criminal Law @chels_bar

Chelsea Barabas, MIT Media Lab, has published Beyond Bias: Re-Imagining the Terms of ‘Ethical AI’ in Criminal Law. Here is the abstract.

Data-driven decision-making regimes, often branded as “artificial intelligence,” are rapidly proliferating across the US criminal justice system as a means of predicting and managing the risk of crime and addressing accusations of discriminatory practices. These data regimes have come under increased scrutiny, as critics point out the myriad ways that they can reproduce or even amplify pre-existing biases in the criminal justice system. This essay examines contemporary debates regarding the use of “artificial intelligence” as a vehicle for criminal justice reform, by closely examining two general approaches to, what has been widely branded as, “algorithmic fairness” in criminal law: 1) the development of formal fairness criteria and accuracy measures that illustrate the trade-offs of different algorithmic interventions and 2) the development of “best practices” and managerialist standards for maintaining a baseline of accuracy, transparency and validity in these systems. The essay argues that attempts to render AI-branded tools more accurate by addressing narrow notions of “bias,” miss the deeper methodological and epistemological issues regarding the fairness of these tools. The key question is whether predictive tools reflect and reinforce punitive practices that drive disparate outcomes, and how data regimes interact with the penal ideology to naturalize these practices. The article concludes by calling for an abolitionist understanding of the role and function of the carceral state, in order to fundamentally reformulate the questions we ask, the way we characterize existing data, and how we identify and fill gaps in existing data regimes of the carceral state.

Download the article from SSRN at the link.