Elon Musk: SpaceX, Mars, Tesla Autopilot, Self-Driving, Robotics, and AI | Lex Fridman Podcast #252

In his third appearance on the Lex Fridman podcast, Elon Musk dives deep into the engineering behind SpaceX and Tesla. He explains why production, not design, is the hardest problem, why Raptor's full-flow staged combustion engine is revolutionary, and why fully reusable rockets are the holy grail for reaching Mars. He details Tesla's approach to autonomy as recreating human vision in silicon, including building an accurate vector space, cutting latency and jitter, and replacing C++ heuristics with neural nets. The conversation ranges across the Tesla Bot, cryptocurrency as an information-theory problem, direct democracy on Mars, World War II aircraft engineering, nuclear power fears, a freewheeling meme review, and reflections on history, love, and human consciousness.

• Musk says engine production, not design, is by far the biggest thing absorbing his time on Starship.
  00:08:22
• SpaceX plans to catch the largest flying object ever made with giant 'Chopstick' tower arms to save the weight of landing legs.
  00:13:41
• Musk estimates landing humans on Mars in about 5 years best case, 10 years worst case.
  00:27:23
• Current cost is roughly a billion dollars per ton to the surface of Mars, which must drop by a factor of a thousand or more.
  00:33:50
• Musk says the global money system literally runs on heterogeneous mainframes executing ancient COBOL in batch mode.
  00:51:42
• Musk predicts Tesla will likely solve level four full self-driving 'next year,' targeting safety 2-3x better than humans.
  01:26:41
• Musk admits he had never thought of the Tesla Bot as a companion, but believes it could become a very good one.
  01:45:39
• Musk says he might have enough material for a 15-minute stand-up comedy set and would consider a Netflix special.
  02:14:27

• Raptor targets 300 bar chamber pressure versus the record ~267 bar of the Russian RD engine, and 10% more pressure is ~50% harder.
  00:09:58
• Musk's 'magic wand number' is the cost if you could rearrange raw material atoms into the final shape, almost always a very low number.
  00:21:36
• Earth becomes too hot for life in about 500 million years, only 10% longer than life has existed.
  00:30:14
• Full and rapid reusability could cut cost per ton to orbit by a factor of 100, with a Starship launch potentially costing $1-2 million.
  00:39:33
• Your eyes paint in color in peripheral vision and fill in the blind spot; the brain does enormous post-processing.
  01:11:01
• Tesla wrote its own C compiler and runs primarily C, not C++, for maximum performance on the autopilot hardware.
  01:18:51
• Tesla moved to raw photon counts instead of post-processed images, saving 13 milliseconds of latency and seeing better in the dark.
  01:21:29
• In WWII, Germany had great aircraft designs but lacked consistent high-octane fuel and quality aluminum alloys, unlike the US.
  01:48:19
• Musk notes coal plants are roughly 100 to 1,000 times worse for health than nuclear power plants.
  02:04:34
• Musk warns against a 'zero-sum mindset' and urges creating more than you consume, since the economic pie is not fixed.
  02:23:20