MachineLearning Podcast Summaries
MachineLearning on Yedapo: 4 summarized podcast and YouTube episodes. Each includes key takeaways, core concepts and notable quotes with timestamps.
AI Just Entered A New Era
Two Minute Papers
Jul 1, 2026
The emergence of GLM 5.2 proves that open-weights AI is rapidly approaching the capabilities of proprietary systems like Claude. By utilizing advanced training techniques like multi-token prediction and individual step-grading, these models offer a future where individuals can own and control powerful, uncensored intelligence rather than relying on restrictive, government-regulated corporate platforms.
Key insight: GLM 5.2 employs an ingenious 'anti-hacking' measure that lets an AI attempt to cheat on benchmarks, only to provide it with false data so it gains no advantage, effectively neutralizing benchmark manipulation.
AI Foundations for Absolute Beginners
freeCodeCamp.org
Mar 26, 2026
While the media pits humans against machines, Henry Lee reveals that AI is fundamentally a product of human values and mathematical patterns. This masterclass strips away the "black box" mystery, demonstrating how to build classifiers while exposing why algorithmic bias starts with our own limited perspectives.
Key insight: AI "learning" is actually a series of calculated predictions where features are turned into numbers and adjusted by mathematical importance—a process so intensive that AI data centers can consume as much electricity as entire small cities.
Deploying AI Models with Hugging Face – Hands-On Course
freeCodeCamp.org
Mar 25, 2026
The gap between AI research and production-ready systems is closing faster than ever. This masterclass reveals how the Hugging Face ecosystem streamlines the entire pipeline—from tokenizing complex linguistics to deploying generative vision systems. Mastering these open-source tools allows developers to build sophisticated multimodal applications in minutes rather than months.
Key insight: GPT-2 doesn't read words like humans; it breaks the longest word in English—'pneumonoultramicroscopicsilicovolcanoconiosis'—into exactly 15 distinct tokens, demonstrating the mathematical precision required for machine comprehension.
Desbloqueando o "Algoritmo" do Twitter - Introdução a Grafos
Fabio Akita
Apr 11, 2023
Fabio Akita analisa o código aberto do Twitter, demonstrando que a recomendação é uma coleção de dezenas de algoritmos complexos, não uma única peça de lógica. Ele alerta que sem os arquivos de build e dados de treino, a análise estática é inconclusiva, desmascarando interpretações superficiais sobre supostos vieses políticos no código.
Key insight: O sistema de recomendação evoluiu em gerações, do 'Who To Follow' (SALSA) para o GraphJet, priorizando a eficiência de processamento em memória para lidar com o volume massivo de interações, provando que escalar sistemas exige matemática avançada, não apenas truques de código.