No Priors Ep. 80 | With Andrej Karpathy from OpenAI and Tesla

Andrej Karpathy joins Sarah and Elad in this week of No Priors. Andrej, who was a founding team member of OpenAI and the former Tesla Autopilot leader, needs no introduction. In this episode, Andrej discusses the evolution of self-driving cars, comparing Tesla's and Waymo’s approaches, and the technical challenges ahead. They also cover Tesla’s Optimus humanoid robot, the bottlenecks of AI development today, and how AI capabilities could be further integrated with human cognition. Andrej shares more about his new mission Eureka Labs and his insights into AI-driven education and what young people should study to prepare for the reality ahead. Sign up for new podcasts every week. Email feedback to show@no-priors.com Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @Karpathy Show Notes: 0:00 Introduction 0:33 Evolution of self-driving cars 2:23 The Tesla vs. Waymo approach to self-driving 6:32 Training Optimus with automotive models 10:26 Reasoning behind the humanoid form factor 13:22 Existing challenges in robotics 16:12 Bottlenecks of AI progress 20:27 Parallels between human cognition and AI models 22:12 Merging human cognition with AI capabilities 27:10 Building high performance small models 30:33 Andrej’s current work in AI-enabled education 36:17 How AI-driven education reshapes knowledge networks and status 41:26 Eureka Labs 42:25 What young people study to prepare for the future

Sarah GuohostElad GilhostAndrej Karpathyguest

Sep 4, 202444mWatch on YouTube ↗

WHAT IT’S REALLY ABOUT

Andrej Karpathy on self-driving, humanoid robots, transformers, and education’s future

Andrej Karpathy discusses how self-driving cars preview the path from AI demos to real-world, globalized products, comparing Tesla’s software-centric approach with Waymo’s hardware-heavy strategy. He explains why transformers are a foundational breakthrough, why the bottlenecks now are data and loss functions, and why synthetic data—if kept diverse—is essential for continued progress. The conversation explores humanoid robotics, including Tesla’s Optimus, transfer from car autonomy to robots, and why humanoids and multi-agent “swarms” of models make sense. Karpathy closes by outlining his new education company Eureka, his vision of AI tutors as an exocortex that empowers “team human,” and why rigorous STEM foundations will matter in both pre- and post-AGI worlds.

IDEAS WORTH REMEMBERING

5 ideas

Demos and real products are separated by a long, messy operational gap.

Karpathy notes Waymo drove him flawlessly a decade ago, yet it took 10 years to become a commercial product—highlighting that moving from a polished demo to robust, global deployment demands years of engineering, edge-case handling, and regulatory work.

Tesla’s self-driving advantage is deployment scale and a software-first strategy.

He argues Tesla has a “software problem” (easier to fix) while Waymo has a “hardware problem” (scaling expensive sensor rigs), and that Tesla’s massive installed base plus vision-only deployment—augmented by rich sensors at training time—is a powerful arbitrage.

Transformers are a fundamental architectural unlock; data and loss are now the real bottlenecks.

Unlike earlier architectures, transformers scale cleanly and act like a general-purpose differentiable computer; most innovation now happens in dataset curation and objective design rather than in model architecture.

Synthetic data is essential but must preserve entropy to avoid silent collapse.

He warns that naive synthetic data pipelines produce narrow, repetitive distributions (e.g., LLMs repeating the same few jokes), so future training must deliberately inject diversity—via personas, varied prompts, and other techniques—to maintain richness.

Humanoid robots benefit from massive transfer from autonomous driving and from a unified platform.

Tesla’s robot reused car hardware, software, and data infrastructure so seamlessly that early Optimus “thought it was a car”; Karpathy believes a single humanoid platform maximizes reuse, teleoperation ease, and cross-task transfer learning.

WORDS WORTH SAVING

5 quotes

I kind of feel like we’ve reached AGI a little bit in self-driving, because there are systems today that you can basically take around as a paying customer.

— Andrej Karpathy

I think Tesla has a software problem, and I think Waymo has a hardware problem, and I think software problems are much easier.

— Andrej Karpathy

The transformer is not just another neural net; it’s a general-purpose training computer.

— Andrej Karpathy

If it’s not your weights, it starts to feel like you’re renting your brain.

— Andrej Karpathy

I feel like I’m kind of, on a high level, team human, and I’m interested in things that AI can do to empower people.

— Andrej Karpathy

State of self-driving cars and Tesla vs. Waymo approachesEnd-to-end deep learning and transformers as a general-purpose “differentiable computer”Humanoid robots, Tesla Optimus, and robotics use cases and strategySynthetic data, dataset design, and risks of model collapseAI cognition vs. human cognition and the exocortex conceptMarket structure: closed vs open-source models and ownership of “your brain”AI for education, Eureka, and the future of learning and human capability

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.