Andrej Karpathy, just 36 years old, has had an extraordinary career in computer science and artificial intelligence. He created and taught the first course in deep learning at Stanford University, which has since become one of the largest courses in the department. In 2015, he joined OpenAI as a founding member and research scientist, and in 2017 left to become director of artificial intelligence at Tesla, in charge of Autopilot Vision for their self-driving vehicle project. In July 2022, he announced he was leaving Tesla. Before coming onto the radar in the field of artificial intelligence and machine learning and vision, he was known as a lightning solver of Rubik’s cubes, creating a YouTube channel on speedcubing techniques.
In this three and a half hour conversation with Lex Fridman, he discusses topics ranging from details of solving the self-driving problem, the applicability of deep neural networks to a variety of problems, the origin of life and intelligence in the universe, the prospects for artificial general intelligence and human civilisation, and Tesla’s humanoid robot project.
Here are chapter marks for topics in the discussion.
0:00 - Introduction
0:58 - Neural networks
6:01 - Biology
11:32 - Aliens
21:43 - Universe
33:34 - Transformers
41:50 - Language models
52:01 - Bots
58:21 - Google’s LaMDA
1:05:44 - Software 2.0
1:16:44 - Human annotation
1:18:41 - Camera vision
1:23:46 - Tesla’s Data Engine
1:27:56 - Tesla Vision
1:34:26 - Elon Musk
1:39:33 - Autonomous driving
1:44:28 - Leaving Tesla
1:49:55 - Tesla’s Optimus
1:59:01 - ImageNet
2:01:40 - Data
2:11:31 - Day in the life
2:24:47 - Best IDE
2:31:53 - arXiv
2:36:23 - Advice for beginners
2:45:40 - Artificial general intelligence
2:59:00 - Movies
3:04:53 - Future of human civilization
3:09:13 - Book recommendations
3:15:21 - Advice for young people
3:17:12 - Future of machine learning
3:24:00 - Meaning of life