Latest Interviews
Showing 1–4 of 4 transcripts.
Clear all filters- Lex Fridman1h 5m
Mark Zuckerberg: First Interview in the Metaverse | Lex Fridman Podcast #398
Meta demonstrated photorealistic 3D avatar technology through a remote conversation between Mark Zuckerberg and Lex Friedman, utilizing efficient codec compression to transmit subtle facial imperfections and spatial audio for enhanced emotional presence. This capability is slated for integration into the upcoming $500 Quest 3 headset, which features advanced mixed reality pass-through and latency-free hand tracking to blend digital avatars with physical environments. Looking forward, the roadmap includes reducing avatar scanning time to minutes via smartphones and deploying AI-generated personas to transform remote work, social connection, and creative interactions within a unified digital-physical reality.
- Lex Fridman9 min
Pixel 6 AI explained | Lex Fridman
Google introduces the Pixel 6 Pro, a flagship device powered by the custom Google Tensor system-on-a-chip designed to optimize real-time neural network tasks and computational photography. The hardware features a 120Hz display, a 48MP telephoto lens with 4x optical zoom, and a heterogeneous CPU configuration that reportedly improves thermal management during intensive video recording compared to competitors. While benchmarks show the Snapdragon 888 leads in general multi-core performance, the Tensor chip secures dominance in machine learning metrics, enabling AI-driven software features like Magic Eraser, Real Tone, and super-resolution zoom.
- Lex Fridman18 min
Comma.ai Drive and Tour with George Hotz and Lex Fridman
George Hotz demonstrated the Comma 2 system's end-to-end neural network capabilities during a San Diego drive, where the vehicle successfully navigated complex banked curves and lane changes without disengagement while leveraging Hyundai's existing longitudinal control. Operating from multinational facilities in the U.S. and Europe, the company has grown its active user base to nearly 3,000 monthly devices and now manages 30 million miles of data to validate safety through rigorous injection testing. With plans to deploy OpenPilot 1.1 via over-the-air updates and integrate new machine learning talent, the firm aims to achieve full holistic autonomy on highways before expanding to dense city environments.
- Lex Fridman1h 3m
Theano Tutorial (Pascal Lamblin, MILA)
Pascal Lamblin, Hugo Larochelle, Andrej Karpathy, Richard Socher, Sherry Moore, Ruslan Salakhutdinov, Andrew Ng, John Schulman, Adam Coates, Alex Wiltschko, Quoc Le, Yoshua Bengio, Shubho Sengupta
This presentation details Theano, an eight-year-old symbolic expression compiler that enables high-performance deep learning by automatically differentiating mathematical graphs and compiling them into optimized C++ or CUDA code. The session demonstrates the framework's core capabilities, including graph manipulation for neural network backpropagation, GPU acceleration via shared variables, and sequence modeling through the `scan` operator, while showcasing practical implementations of logistic regression, convolutional networks, and LSTMs on datasets like MNIST. Addressing deployment challenges inherent in its tight Python integration, the discussion concludes by highlighting Docker containers as the standard solution for distributing models and outlines a roadmap for enhanced 3D convolution and cuDNN support.