Welcome to Machine Culture’s “Model Convo” series — micro interviews with researchers in AI policy and related fields. If you know someone I should interview, send them my way! I’m also taking guest contributors for the “Other Minds” segment. So, if you have a pitch, send that too. We offer small stipends for contributors.
Name: Zandi Eberstadt.
Present: AI policy researcher at the Mercatus Center; author of Corpus Mechanica; incoming computer science doctoral candidate at the University of Oxford; appreciator of friends, family, caffeine, meditation, exonyms, and the global etymologies of “tea.” (TLDR: We can thank the Silk Road.)
Past: Computational linguistics and cognitive science at Johns Hopkins University; machine learning and forensic authorship attribution at the Université Paris Sciences et Lettres; winner of the Cosmos Institute’s inaugural essay competition; actually worked in fashion modelling.
How did you get into AI?
Early in college, I was really (honestly, inordinately) attached to theoretical linguistics and particularly syntax trees. Perhaps amidst the chaos of life, I was very drawn to an orderly, predictable blueprint underpinning human language. The universality of Noam Chomsky’s theory of grammar was also very appealing to me. I think that I always search for a “theory of everything,” whether in science, the Bible, or Traditional Chinese Medicine.
But then, when I was a junior in college, ChatGPT exploded and seemingly shattered some of my staunchest beliefs about language. I saw that grammatical sentences could be assembled stochastically, without appealing to syntax trees! Further, it did not seem that the ability to generate grammatical language made humans distinct from all other forms of matter! From there, I became obsessed with learning more about the nature of language and what it means to be human through a lens of math and science.
What work of art has most shaped your views on emerging tech?
Oh, gosh. As a personal side project, I’ve been trying to read a work by every Nobel Prize in Literature laureate, which has embarrassingly devolved into a “hiatus” the past few months, meaning a pile of lofty intentions and unread paperbacks.
Kazuo Ishiguro’s book, Klara and the Sun, is about a robot companion, and – without spoiling the end – it offers a really thought-provoking solution to what it means to have a self, soul, or personhood. Brilliant.
What’s your most contrarian take on AI?
In the scope of things, language will be a transitional phenomenon. I feel deeply attached to language, but both text and speech are painfully inefficient — they take forever to get across — and they are lossy — thoughts need to be linearized, filtered through grammar, etc. Language is computationally expensive to learn, only for the communicator to then be subjected to huge amounts of misunderstanding.
In the next 500 years — assuming our species will still be around, which is by no means guaranteed — I think that our descendants and AI hybrids will transition towards a post-linguistic space, where pure bit-encoded information or cognitive states will be transmissible more quickly and accurately than they are now, agent to agent, mind to mind.
What are you reading now?
Kai-Fu Lee and Chen Qiufan’s AI2041 — these are short stories that forecast where AI could take us in the next two decades. The prose is sharp as a razor, touching on virtual reality, augmented reality, extended reality, educational technology, celebritydom, and more.
Let’s hear your go-to emerging tech policy track.
“Dreamin, Dreamin” by CAPSULE — because nothing can power you through a rapidly changing field like synths and 140 beats per minute.