AI Summary
5 min read“It saw the pedestrian through the bus.” That is how Waymo co-CEO Dmitri Dolgov describes a moment that still surprises him. In a San Francisco intersection, a Waymo vehicle’s light turned green, and it began nudging around a bus that was partially blocking the lane. The car then slowed and gave extra space—before a pedestrian emerged from the other side of the bus. The sensors hadn’t seen through the metal. Instead, a peripheral lidar beam had bounced under the bus, catching a faint, noisy reflection of the pedestrian’s moving feet. The AI model had enough signal to infer a person was there and predict their trajectory. For Dolgov, this was a concrete example of how a system built with multiple sensing modalities and learned intermediate representations can do things that look almost magical—but are actually the result of careful architecture.
The Architecture: Foundation Model, Three Teachers, and Distillation
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of a16z Show
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (00:00) **Episode Introduction** - Host sets up the conversation with Waymo co-CEO Dmitri Dolgov, framing the shift from long-term research to real-world scale.
- 2 (01:41) **Dmitri’s Background & Russian Math-Science Pedigree** - Dolgov describes growing up in the Soviet Union and his family’s path to the US, highlighting the strong foundational education he received.
- 3 (04:26) **The Real-Time Architecture of a Waymo Ride** - Dolgov explains the core technical stack that processes sensor data and actuates the vehicle, all done locally on board.
- 4 (07:52) **End-to-End vs. Modular: The Real Technical Debate** - Dolgov addresses the common Twitter debates, arguing they lose crucial nuance, and reveals Waymo’s actual three-teacher architecture.
- 5 (11:19) **The Enabling Technology: Not Just Scaling Laws** - Dolgov explains that while compute and AI breakthroughs (like transformers) were critical, the journey was iterative learning, not a series of dead ends.
- 6 (17:38) **Why Intermediate Representations are Necessary** - Dolgov explains that to handle the long tail of edge cases, the system must augment learned embeddings with structured representations of the world.
- 7 (21:06) **The Nuances of Driving: Safety, Smoothness, and Social Behavior** - Dolgov details the multi-faceted optimization problem, including safety, ride comfort, and social predictability, as well as the challenge of drop-offs.
+ Full timestamped outline available in the app
Show Notes
Waymo is now delivering hundreds of thousands of fully autonomous rides each week — but getting there required more than better models. It meant building a complete system for training, evaluating, and deploying a driver in the real world.
In this episode — originally aired on the Cheeky Pint podcast — Waymo Co-CEO Dmitri Dolgov joins John Collison to break down how self-driving actually works today: from sensor fusion across LiDAR, radar, and cameras, to simulation, “critic” models, and the role of AI in decision-making.
They also explore why full autonomy is fundamentally different from driver-assist, what it takes to scale globally, and how recent advances in AI are reshaping the path forward.
Resources:
Follow Dmitri Dolgov on X - https://x.com/dmitri_dolgov
Follow John Collison on X - https://x.com/collision
Stay Updated:
Find a16z on YouTube: YouTube
Find a16z on X
Find a16z on LinkedIn
Listen to the a16z Show on Spotify
Listen to the a16z Show on Apple Podcasts
Follow our host: https://twitter.com/eriktorenberg
Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures.
Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.
More from this podcast
a16z Show →