By Julian Adler, Jethro Antoine, Laith Al-Saadoon 

There was but one passing reference to “core values” over the course of a recent U.S. Senate Judiciary hearing on artificial intelligence [AI] in criminal investigations and prosecutions.[1] This is typical. Even in spaces like the criminal legal system, where the specters of racial injustice and inhumanity loom so large, the technological sublimity of AI can be awfully distracting. People have long looked to technology to duck the hard problem of values. “[W]e have tended to believe that if we just had more information, we could make better policy,” observes University of Nevada’s Lynda Walsh in Scientists as Prophets. “But no matter how much data we could lay hands to—even if it were LaPlace’s Demon itself—values would still stand in the way.”[2] If anything is clear about advanced AI, it is that there is much we don’t know and even more that we can’t begin to predict. Consider that the “generative AI” we have witnessed over the past 18 months—AI which produces autonomous human-impersonating content—was largely unforeseen. It’s now being attributed to AI’s “emergent abilities.”[3] Across sectors, most observers acknowledge that AI is a game-changing technology. The Financial Industry Regulatory Authority is illustrative: using AI, it now processes “a peak volume of 600 billion transactions every day to detect potential abuses,” making the regulator “one of the largest data processors in the world.”[4] Tell  ingly, many of the people closest to the leading edges of AI development are sounding the loudest alarms about its capabilities. “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war,” warned the Center for AI Safety in 2023.[5] AI has the potential to supercharge, not mitigate, the uglier sides of humanity, much like, as one journalist puts it, “a fun-house-style… mirror magnifying biases and stripping out the context from which their information comes.”[6] Advanced AI is “not just another technology,” contends Nick Bostrom, Director of the Future of Humanity Institute at the University of Oxford. It is not “another tool that will add incrementally to human capabilities.”[7] Echoing countless dystopian projections of the future, the Center for AI Safety predicts AI systems will likely “become harder to control” than previous forms of technology; among other disquieting scenarios, these systems could “drift from their original goals” and “optimize flawed objectives.”[8] 

New York: Center for Court Innovation, 2024. 8p.