Nicole Williams Self Taught Student on AI, Online Learning, Fashion

About this episode

Nicole Williams, a 22-year-old self-taught VC investor at seed-stage tech firm Compound, joins Vance for a wide-ranging conversation spanning self-directed education, Twitter culture, AI ethics, fashion, religion, and farm data privacy. Williams describes an unconventional path — one semester at Liberty University, a software research job, a medical-tech accelerator, six months at Lambda School, then venture capital — built on self-curated "syllabi" assembled from friends' coursework rather than formal credentials. The conversation's centerpiece is her concept of "algorithmic populism": recommendation systems (Amazon, urban/airport design) flatten cultural taste toward the statistical middle, eroding the "boutique" diversity of experience once produced by less-curated markets. She reframes "radicalization" not as extremity of belief but as unwillingness to update beliefs when challenged — a lens Vance applies to his own controversial mask tweet. A substantial segment covers AI ethics: the is/ought problem in algorithmic decision-making, facial recognition's dual use in surveillance (citing Uyghur camps) and security research, GPT-2's unsettling essay-writing capability, and "federated learning" as a privacy-preserving alternative to raw data harvesting (Google Keyboard, Apple/Stanford heart research). The two also explore religion and culture: Williams's evangelical Christian upbringing at Liberty, the historical is/faith tension dating to 13th-century Aristotelian debates, and Vance's Peace Corps-informed observations on monochronic vs. polychronic time applied to churchgoing. The episode closes on agricultural data trust — Vance argues ag-tech VCs need direct farmer relationships to earn the trust required for data-sharing adoption — and predictions about post-COVID remote work enabling geographic dispersal away from coastal tech hubs.

Key moments

• **[00:17:49]** Introduction of "algorithmic populism" — the thesis that algorithmically curated experiences (retail, design, taste) flatten diversity toward the statistical middle.
• **[00:14:31]** Redefining "radicalization" as unwillingness to hear other perspectives rather than extremity of belief, applied directly to Vance's own controversial mask tweet from the prior day.
• **[00:45:00]** AI ethics discussion: historical bias "baked into" predictive algorithms (insurance, legal systems) being projected forward as if neutral and data-backed.
• **[00:50:35]** Explanation of federated learning as a privacy-preserving alternative to full data harvesting, with Google Keyboard and Apple/Stanford heart-disease research as examples.
• **[00:55:15]** Farm-data trust dilemma: ag-tech companies collecting granular yield/planting data risk market misuse (futures trading) unless farmers can trust the data recipient; Williams recommends farmer-led transparency organizing.