Dec 21, 2019, 12:49 PM ETInterview

Sebastian Thrun: Flying Cars, Autonomous Vehicles, and Education | Lex Fridman Podcast #59

Thrun considers the "simulation hypothesis" irrelevant to human behavior, prioritizing the present moment and human connection over abstract theoretical physics.
He views the universe as a massive, unpredictable, information-processing physical, biological, and chemical computer without a specific intentional goal.
He believes the most significant innovation in AI history is machine learning, allowing systems to derive their own rules from data rather than relying on rigid, hand-crafted human instructions.
Thrun argues that human expertise often resides in subconscious, perceptual, and numerical data that cannot be articulated through language or written rules.
AI systems can now surpass human capabilities by observing experts (e.g., doctors, lawyers) and extracting implicit skills that humans cannot consciously describe.

Thrun identifies two core life drivers: making the world a better place and constantly learning by deliberately entering domains where he lacks expertise.
He advocates for a funding model shift in academia from "effort-based" (paying for hours worked/papers written) to "outcome-based" (paying for solved problems), a strategy proven successful by the DARPA Grand Challenge.
The 2005 DARPA Grand Challenge forced a reset in robotics, moving focus from incremental academic papers to building functional, end-to-end systems that could solve real-world problems.
Thrun emphasizes that the best leaders do not treat teams like computers but instead empower employees by connecting to their intrinsic good intentions and desire to contribute.
He credits his success at Stanford and with the DARPA team to extreme focus on "time management," freezing software early to allow a month of rigorous stress-testing and bug fixing before deadlines.
He notes that 90% of self-driving car development is easy; the remaining 10% involves solving the long-tail edge cases that account for the vast majority of safety risks.

Thrun observes a industry-wide shift from geometric reasoning (defining rules for lane markers) to deep learning (training models on massive datasets of human driving).
He validates the "eyes-only" approach for autonomy, citing the existence proof that humans drive perfectly with only visual input, supporting Tesla's reliance on cameras.
He predicts that Level 4/5 autonomy without safety drivers is the next major hurdle, noting that while cars like Waymo's operate flawlessly with drivers present, empty-vehicle traffic remains the "magical hurdle" to clear.
Thrun estimates the value of a skilled self-driving engineer at over $10 million based on recent acquihires, highlighting the massive economic impact of democratizing this skill.
He believes autonomous vehicles will eventually save 1.2 million lives annually and drastically reduce traffic congestion, transforming the social fabric of cities.
He defends the US innovation model as an "anthill" strategy where diverse, decentralized approaches (like Tesla's aggressive iteration vs. Waymo's safety-first approach) allow society to find the best solution naturally.

Thrun cites a Stanford study where an AI system diagnosed skin cancer as accurately as board-certified dermatologists, demonstrating AI's potential to detect life-threatening conditions early.
He envisions AI assisting doctors in their first 10,000 hours of practice, effectively making every doctor an "expert" on their first day by leveraging AI pattern recognition.
Thrun predicts AI will not replace humans but will eradicate the lack of expertise and decision-making errors in professions like medicine and law.
He addresses job displacement fears by promoting massive upskilling initiatives, such as Udacity's 100,000 scholarships for US citizens to gain tech skills.
Thrun believes education must be treated as a basic human right, accessible globally regardless of geography, demographics, or age.
He identifies "soft skills" (empathy, teamwork, management) as the next critical frontier for education, noting that universities currently fail to teach the interpersonal skills required for modern leadership.

Thrun's company, Kittyhawk, focuses on Electric Vertical Takeoff and Landing (eVTOL) aircraft, specifically highlighting the "Project Heaviside" which operates at 38 decibels (library quiet) and features redundant motor systems for safety.
He argues that eVTOLs are safer than cars because the sky is a three-dimensional space where traffic lanes are virtual and can be managed by software, unlike the congested 1D plane of roads.
Thrun predicts eVTOLs will be 100% autonomous, as human pilots are unnecessary risks in an environment where digital air traffic control can manage thousands of vehicles simultaneously.
He estimates that widespread adoption of eVTOLs could reduce annual commute time from 300 hours to 30 hours, effectively freeing up 270 hours for human use.
He asserts that the technical challenges for eVTOLs are solved, and the primary remaining barriers are societal acceptance regarding noise and safety regulations.

Thrun rejects the idea of machines possessing human-like emotions or "love," viewing technology as a tool designed for reliability, predictability, and augmentation rather than emotional reciprocity.
He argues that engineering solutions (like wheels) often differ fundamentally from biological ones (which rarely evolve rotating wheels), suggesting we should not blindly mimic nature for technological advancement.
He attributes his optimism and humor to a profound appreciation for the current era, noting that modern life expectancy and technological abundance are historical anomalies worth celebrating.
Thrun references Carl Bosch's invention of nitrogen fertilization as a singular innovation that saved over 2 billion lives, using it to illustrate the massive, often unrecognized impact of human engineering.
He concludes that the most effective way to move the world is to celebrate failure as a learning opportunity, removing the fear that paralyzes innovation.