Paul Smaldino & C. Thi Nguyen on Problems with Value Metrics & Governance at Scale (EPE 06)
COMPLEXITYShow Notes
There are maps, and there are territories, and humans frequently confuse the two. No matter how insistently this point has been made by cognitive neuroscience, epistemology, economics, and a score of other disciplines, one common human error is to act as if we know what we should measure, and that what we measure is what matters. But what we value doesn’t even always have a metric. And even reasonable proxies can distort our understanding of and behavior in the world we want to navigate. Even carefully collected biometric data can occlude the other factors that determine health, or can oversimplify a nuanced conversation on the plural and contextual dimensions of health, transforming goals like functional fitness into something easier to quantify but far less useful. This philosophical conundrum magnifies when we consider governance at scales beyond those at which Homo sapiens evolved to grasp intuitively: What should we count to wisely operate a nation-state? How do we practice social science in a way that can inform new, smarter species of political economy? And how can we escape the seductive but false clarity of systems that rain information but do not enhance collective wisdom? Welcome to COMPLEXITY, the official podcast of the Santa Fe Institute. I’m your host, Michael Garfield, and every other week we’ll bring you with us for far-ranging conversations with our worldwide network of rigorous researchers developing new frameworks to explain the deepest mysteries of the universe. This week on the show we talk to SFI External Professor Paul Smaldino at UC Merced and University of Utah Professor of Philosophy C. Thi Nguyen. In this episode we talk about value capture and legibility, viewpoint diversity, issues that plague big governments, and expert identification problems…and map the challenges “ahead of us” as SFI continues as the hub of a five-year international research collaboration into emergent political economies. (Find links to all previous episodes in this sub-series in the notes below.) Be sure to check out our extensive show notes with links to all our references at complexity.simplecast.com. If you value our research and communication efforts, please subscribe, rate and review us at Apple Podcasts or Spotify, and consider making a donation — or finding other ways to engage with us — at santafe.edu/engage. If you’d like some HD virtual backgrounds of the SFI campus to use on video calls and a chance to win a signed copy of one of our books from the SFI Press, help us improve our science communication by completing a survey about our various scicomm channels. Thanks for your time! Lastly, we have a bevy of summer programs coming up! Join us June 19-23 for Collective Intelligence: Foundations + Radical Ideas, a first-ever event open to both academics and professionals, with sessions on adaptive matter, animal groups, brains, AI, teams, and more. Space is limited! The application deadline has been extended to March 1st. OR apply to the Graduate Workshop on Complexity in Social Science. OR the Complex ity GAINS UK program for PhD students. (OR check our open listings for a staff or research job!) Join our Facebook discussion group to meet like minds and talk about each episode. Podcast theme music by Mitch Mignano. Follow us on social media: Twitter • YouTube • Facebook • Instagram • LinkedIn Mentioned & Related Links: Transparency Is Surveillance by C. Thi Nguyen The Seductions of Clarity by C. Thi Nguyen The Natural Selection of Bad Science by Paul Smaldino and Richard McElreath Maintaining transient diversity is a general principle for improving collective problem solving by Paul Smaldino, Cody Moser, Alejandro Pérez Velilla, Mikkel Werling The Division of Cognitive Labor by Philip Kitcher The Unreasonable Effectiveness of Mathematics in The Natural Sciences by Eugene Wigner On Crashing The Barrier of Meaning in A.I. by Melanie Mitchell Seeing Like A State by James C. Scott Jim Rutt Slowed Canonical Progress in Lar
More from COMPLEXITY
In the final episode of the season, Abha sits down with Melanie to hear her perspective. They chat about Melanie’s career and research with Douglas Hofstadter, the author of Gödel, Escher, Bach. They also discuss her opinions on LLMs’ current capabilities, what she thinks of existential questions like the alignment problem, how sustainable the industry is, the difficulty of making claims about concepts like “intelligence” and “understanding,” and what she thinks future technological development should focus on.
When it comes to assessing intelligence, people have all kinds of tests — the SAT, IQ tests, and so on. There’s controversy over how fairly these tests really measure human intelligence, but at the very least, we know that they correlate with some general reasoning skills when people take them. That assumption breaks down when we try to assess intelligence in non-humans. What does it mean when a large language model passes an intelligence test meant for humans? Does it actually have the same reasoning skills that a human does, or is it doing something else? In today’s episode, with guests Erica Cartmill and Ellie Pavlick, we investigate the best ways to assess intelligence in non-humans, whether animals or machines.
There’s an argument to be made that if we train AI systems to learn the way babies do, we’ll get them closer to human-like intelligence. But how our own learning development functions in babyhood is still a mystery that researchers are untangling. We know that the information babies absorb is very different from how an LLM learns, and in today’s episode, with guests Linda Smith and Michael Frank we’ll attempt to look at the world through an infant’s eyes and examine why they’re able to do more with, seemingly, less information.
Large language models, like ChatGPT and Claude, have remarkably coherent communication skills. Yet, what this says about their “intelligence” isn’t clear. Is it possible that they could arrive at the same level of intelligence as humans without taking the same evolutionary or learning path to get there? Or, if they’re not on a path to human-level intelligence, where are they now and where will they end up? In this episode, with guests Tomer Ullman and Murray Shanahan, we look at how large language models function and examine differing views on how sophisticated they are and where they might be going.
Complex language is unique to the human species. It’s part of how we evolved, the backbone of our societies, and one of the primary ways we judge others’ intellect. Is it our intelligence that leads to our language abilities, or conversely, does our ability for language enhance our intelligence, or both? How do language and thinking interact? And can one exist without the other? Guests: Evelina Federenko, Steve Piantadosi, and Gary Lupyan.