Interview
Kyle Vogt: Cruise Automation | Lex Fridman Podcast #14
- Kyle Vogt is the President and CTO of Cruise Automation, a company acquired by General Motors (GM) for $1 billion, aiming to solve vehicle automation challenges.
- Vogt is a co-founder of two companies, Twitch and Cruise, both of which were sold for billion-dollar valuations.
- Vogt grew up in Kansas and was drawn to robotics by BattleBots, which he viewed as a mechanical engineering challenge involving pushing components like $40 DC motors to their limits by running them at double or triple their rated voltage.
- His first programming experience involved learning BASIC on Apple II computers in the fifth grade, where he appreciated the "magic" of code opening up endless possibilities.
- Vogt's interest in autonomous vehicles sparked during a 10-hour drive across Western Kansas in his early teens, where he realized a computer could theoretically handle the monotonous task of highway driving.
- Prior to the deep learning wave, Vogt's approach to autonomous driving relied on heuristic-based image processing, such as color thresholding for lane markers, because computational power was insufficient for neural networks.
- Vogt left MIT during his junior year in 2004 to co-found Justin.tv (later Twitch) after deciding the opportunity to build something meaningful outweighed the risk of dropping out, despite few peers at the time making the same move to California.
- On March 18, 2007, Vogt and his team faced a critical failure just hours before Justin.tv's public launch, where the video streaming system failed; Vogt spent the night "hacking through the dark" to implement complex error correction and packet reassembly over unreliable cellular networks.
- In October 2013, Vogt founded Cruise Automation with a three-part thesis: the work must involve hard technology problems, have a direct positive impact on society (e.g., saving lives), and address a massive business opportunity.
- Cruise's initial strategy was to sell retrofit kits for highway driving on specific Audi models to generate revenue for fully driverless car research, but the company abandoned this approach after realizing the safety liability and logistical complexity of supporting 20–50 vehicle models.
- The acquisition by GM in 2016 created a significant cultural clash between Silicon Valley's experimental, risk-tolerant methodology and GM's process-driven, risk-averse manufacturing culture focused on on-time and on-budget delivery.
- Vogt notes that GM's scale allows Cruise to leverage existing supply chains and assembly processes, avoiding the need to reinvent vehicle manufacturing, which would make the problem of deploying autonomous fleets intractable.
- The primary economic variables for autonomous vehicle viability are vehicle build cost (including sensors), vehicle lifetime (aiming for millions of miles versus 100,000), and utilization rates (hours of operation per day).
- Vogt identifies ride-sharing and delivery (food, groceries, parcels) as the immediate monetization opportunities once the technology achieves a baseline of safety and reliability superior to human driving.
- Cruise plans to offer customization in driving style (e.g., comfort vs. aggressiveness) but will strictly prohibit users from toggling safety risk parameters, as the system must remain safer than a human driver.
- The current core challenge for Cruise is not solving specific scenarios like left turns or construction zones, but rather addressing the "long tail" of edge cases and continuous performance improvements required to consistently outperform human driving standards.
- Vogt rejects the "2050" timeline for mass autonomy, predicting that hundreds of thousands of autonomous vehicles will be operational in major cities in less than five years as the company expands its operating domain to include inclement weather.
- Vogt attributes startup success to three factors: deep passion for the core technology, surrounding oneself with high-integrity and logical co-founders, and the persistence to grind through unsexy operational tasks.
- Vogt views the 2007–2014 period working on gaming content at Twitch as an "existential detour" that ultimately clarified his desire to work on problems with profound societal impact, fueling his decision to launch Cruise.
- For 2019, Cruise's primary objective is crossing the threshold from prototype to production by refining thousands of system components to achieve a "superhuman" level of software performance capable of launching a commercial product.