#9 - Brian Chau - What Next For AI
Impossible d'ajouter des articles
Échec de l’élimination de la liste d'envies.
Impossible de suivre le podcast
Impossible de ne plus suivre le podcast
-
Lu par :
-
De :
À propos de ce contenu audio
In Episode 9 of The Quantum Podcast, Christopher Smith sits down with machine-learning researcher and writer Brian Chau for a fast-moving exploration of how modern AI really works — from competitive programming and pure math, to transformers, quantization, and the hidden mechanics behind today's large language models.
Brian breaks down attention, zero-shot completion, training vs post-training, and how curated datasets transformed basic next-token predictors into conversational systems like ChatGPT. He also shares his path from Canadian and US programming Olympiads to random graph theory at Waterloo, early AI startups, and building Alliance for the Future in response to US regulatory pressures.
The conversation expands into AI hype cycles, stagnation, and why the "Terminator model" of AGI misses the real story. Chris and Brian examine Peter Thiel's warnings about technological slowdown, ideology as "frozen weights" in science, and what it would take for societies to regain their learning rate. They connect AI to cryptography, quantum threats, and governance — exploring whether decentralized networks can react as quickly as centralized systems, and why Network School may be a template for faster, more resilient social structures.
A dense and philosophical episode about AI, progress, and the future of human coordination.
🔗 Follow the guest
Brian Chau
https://x.com/BrianChau57
Follow Christopher Smith
https://x.com/YuviLightman
Quantus Network
https://www.quantus.com/
https://t.me/quantusnetwork
https://linktr.ee/quantusnetwork
⏱ Chapters
00:00 — Competitive programming, math Olympiads & falling into ML
04:00 — Pure math, random graph theory & early AI startups
08:00 — Quantization, floats & efficiency in model design
12:00 — Transformers, attention & the 2017 breakthrough
18:00 — Zero-shot completion & how ChatGPT learned Q&A
24:00 — Post-training, RLHF & the "polite AI" persona
30:00 — AI winters, hype cycles & diminishing returns
36:00 — Peter Thiel, stagnation & the meaning of progress
42:00 — Ideology, "settled science" & frozen weights in institutions
48:00 — Centralization, blockchains & quantum-era security
52:00 — Network states & the learning rate of civilizations
🎥 The Quantus Podcast — Network School Series (Sept 2025)
Network School brings together remote workers, digital nomads, creators, personal trainers, self-improvers, event organizers, and engineers — all collaborating to build themselves up while building the next generation of startup societies. Founded by Balaji Srinivasan — https://x.com/balajis
📍 Recorded Sept 2025 in the Quantus popup studio at Network School, Forest City
Vous êtes membre Amazon Prime ?
Bénéficiez automatiquement de 2 livres audio offerts.Bonne écoute !