Is there value in arguing both sides?
Was Russia’s Annexation of Crimea Legitimate? A Study in Artificial Intelligence
This study used Copilot and Gab AI, two tools of artificial intelligence (AI), to examine the question of whether Russia’s annexation of Crimea was legitimate. Both chatbots were asked to write a brief essay summarizing the history of Crimea, with emphasis on its annexation by Russia. They were then asked to write a two-part essay providing arguments for both sides of the legitimacy issue. This methodology can be used for any number of research projects in economics, law, history, sociology, philosophy, political science and ethics, to name a few. Professors can utilize this methodology to stimulate class discussion. Graduate students can use it to generate initial outlines for their theses and dissertations. It can be used as a starting point for further discussion.
Who wins?
https://scholar.law.colorado.edu/cgi/viewcontent.cgi?article=2616&context=faculty-articles
Risky Speech Systems: Tort Liability for AI-Generated Illegal Speech
How should we think about liability when AI systems generate illegal speech? The Journal of Free Speech Law, a peer-edited journal, ran a topical 2023 symposium on Artificial Intelligence and Speech that is a must-read. This JOT addresses two symposium pieces that take particularly interesting and interlocking approaches to the question of liability for AI-generated content: Jane Bambauer’s Negligent AI Speech: Some Thoughts about Duty, and Nina Brown’s Bots Behaving Badly: A Products Liability Approach to Chatbot-Generated Defamation. These articles evidence how the law constructs technology: the diverse tools in the legal sensemaking toolkit that are important to pull out every time somebody shouts “disruption!”
Each author offers a cogent discussion of possible legal frameworks for liability, moving beyond debates about First Amendment coverage of AI speech to imagine how substantive tort law will work. While these are not strictly speaking First Amendment pieces, exploring the application of liability rules for AI is important, even crucial, for understanding how courts might shape First Amendment law. First Amendment doctrine often hinges on the laws to which it is applied. By focusing on substantive tort law, Bambauer and Brown take the as-yet largely abstract First Amendment conversation to a much welcomed pragmatic yet creative place.
What makes these two articles stand out is that they each address AI-generated speech that is illegal—that is, speech that is or should be unprotected by the First Amendment, even if First Amendment coverage extends to AI-generated content. Bambauer talks about speech that physically hurts people, a category around which courts have been conducting free-speech line-drawing for decades; Brown talks about defamation, which is a historically unprotected category of speech. While a number of scholars have discussed whether the First Amendment covers AI-generated speech, until this symposium there was little discussion of how the doctrine might adapt to handle liability for content that’s clearly unprotected.
Judge AI is coming.
https://yjolt.org/sites/default/files/avery_abril_delriego_26yalejltech64.pdf
ChatGPT, Esq.: Recasting Unauthorized Practice of Law in the Era of Generative AI
In March of 2023, OpenAI released GPT-4, an autoregressive language model that uses deep learning to produce text. GPT-4 has unprecedented ability to practice law: drafting briefs and memos, plotting litigation strategy, and providing general legal advice. However, scholars and practitioners have yet to unpack the implications of large language models, such as GPT-4, for long-standing bar association rules on the unauthorized practice of law (“UPL”). The intersection of large language models with UPL raises manifold issues, including those pertaining to important and developing jurisprudence on free speech, antitrust, occupational licensing, and the inherent-powers doctrine. How the intersection is navigated, moreover, is of vital importance in the durative struggle for access to justice, and low-income individuals will be disproportionately impacted.
In this Article, we offer a recommendation that is both attuned to technological advances and avoids the extremes that have characterized the past decades of the UPL debate. Rather than abandon UPL rules, and rather than leave them undisturbed, we propose that they be recast as primarily regulation of entity-type claims. Through this recasting, bar associations can retain their role as the ultimate determiners of “lawyer” and “attorney” classifications while allowing nonlawyers, including the AI-powered entities that have emerged in recent years, to provide legal services—save for a narrow and clearly defined subset. Although this recommendation is novel, it is easy to implement, comes with few downsides, and would further the twin UPL aims of competency and ethicality better than traditional UPL enforcement. Legal technology companies would be freed from operating in a legal gray area; states would no longer have to create elaborate UPLavoiding mechanisms, such as Utah’s “legal sandbox”; consumers—both individuals and companies—would benefit from better and cheaper legal services; and the dismantling of access-to-justice barriers would finally be possible. Moreover, the clouds of free speech and antitrust challenges that are massing above current UPL rules would dissipate, and bar associations would be able to focus on fulfilling their already established UPL-related aims.
Oops! What should we try next?
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4721955
Police Technology Experiments
Police departments often adopt new surveillance technologies that make mistakes, produce unintended effects, or harbor unforeseen problems. Sometimes the police try a new surveillance technology and later abandon it -either from a lack of success, community resistance, or both. Critics have identified many problems with these tools: racial bias, privacy violations, opacity, secrecy, and undue corporate influence, to name a few. A different framework is needed. This essay considers the growing use of these algorithmic surveillance technologies and argues that they function as technology experiments on human subjects. Such technology experiments result in police reliance on automated systems to engage in investigative stops and consensual encounters, or to increase police presence and surveillance in a community. Not only do these tools act as experiments, in practice they often function as poorly designed and executed experiments on human subjects. Moreover, ethical considerations that are common in the conventional human subjects research context are entirely absent, even though the new technologies involve uncontrolled experiments on people. And because these algorithmic surveillance technologies are often adopted in low-income, communities of color, they function as poorly designed experiments that raise particularly sensitive concerns about ethics and experimentation borne out by historical experience. By understanding the adoption of new algorithmic surveillance tools as experiments on human subjects, we can develop prospective controls and methods of evaluation for the use of these tools by police, ones that balance innovation with ethical responsibility as artificial intelligence becomes a normal part of police investigations.
No comments:
Post a Comment