#4: The truth behind self-driving cars - Missy Cummings

Dec 17, 2024

In this conversation, Dr. Missy Cummings, one of the U.S. Navy’s first female fighter pilots and director at George Mason University's Autonomy and Robotics Center, delves into the truths about self-driving cars. She shares insights on how these vehicles actually function and their significant limitations, particularly in unpredictable environments. Cummings discusses the importance of integrating human oversight, the challenges of collecting quality training data, and the misconceptions surrounding full autonomy. Prepare for a fascinating look into the future of transportation!

Ask episode

AI Snips

Chapters

Transcript

Episode notes

ANECDOTE

From Fighter Pilot to Autonomy Expert

Missy Cummings's experience as a fighter pilot showed how poor human-machine interaction can be deadly.
This led her to study autonomous systems, focusing on how design impacts human interaction.

INSIGHT

Planes vs. Cars in Autonomy

Counterintuitively, creating autonomous planes is easier than cars.
The third dimension of height offers additional safety margins in the air.

INSIGHT

The Training Data Challenge

Self-driving cars rely heavily on training data to perceive and react.
Real-world scenarios and edge cases present challenges due to the sheer volume and variability of data needed.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

We sat down with Dr. Missy Cummings, who served as one of the U.S. Navy’s first female fighter pilots and now works as the director of George Mason University’s Autonomy and Robotics Center. Cummings, an engineer, has been studying autonomous systems for years – today, we break down the autonomy challenge in self-driving cars: how they work, their enormous limitations and a path to a feasible self-driving future.

Episode Links:

Lidar breaks down in the rain: https://www.mdpi.com/1424-8220/24/10/2997?utm_source=www.thedeepview.co&utm_medium=referral&utm_campaign=openai-launches-o1-unveils-200-subscription-tier
George Mason’s Autonomy and Robotics Center: https://marc.gmu.edu/
Waymo’s self-driving statistics: https://www.thedeepview.co/p/the-complicated-statistics-behind-safe-self-driving-cars
Waymo’s steady expansion: https://waymo.com/blog/2024/12/next-stop-miami/
All the investigations into Tesla’s systems: https://www.cnbc.com/2024/12/09/tesla-accused-of-fraudulent-misrepresentation-of-autopilot-in-crash-.html
A breakdown of why AI confabulates: https://mitsloanedtech.mit.edu/ai/basics/addressing-ai-hallucinations-and-bias/

Outline:
1:39 – From airplanes to AVs
3:57 – What makes self-driving cars tick
9:00 – The risk of edge cases
12:49 – The hallucination problem
18:36 – How do we verify if a self-driving car is safe, in a safe way?
24:02 – Waymo and the scaling of safety data
29:40 – The hardware behind self-driving
31:33 – Limitations of Lidar
37:57 – The misconceptions of self-driving cars
40:16 – A self-driving future

Get The Deep View — your daily source for in-depth, fact-based reporting on artificial intelligence — in your inbox every morning. Subscribe here!

Connect with us on X, TikTok and Instagram.

Artificial intelligence is a complicated topic, bound by a number of complex threads — technical science, ethics, safety, regulation, neuroscience, psychology, philosophy, investment and — above all — humanity. On The Deep View: Conversations, Ian Krietzberg, host and Editor-in-Chief at The Deep View, breaks it all down, cutting through the hype to make clear what's important and why you should care.