Chapter
The Future of Autonomous Vehicle Technology
1:48:47 - 2:00:11 (11:23)
The field of autonomous vehicle technology has been evolving in end-to-end models that solve behavior prediction, decision-making, planning, and detection classification jointly. The result is powerful models that can detect and classify objects in autonomous vehicles and estimate stable stakes accurately.
Clips
The trend in autonomous driving is moving towards end-to-end models that are solving tasks jointly through fusion of sensing data from different modalities at the temporal level rather than using a frame-by-frame detection method.
1:48:47 - 1:52:46 (03:59)
Summary
The trend in autonomous driving is moving towards end-to-end models that are solving tasks jointly through fusion of sensing data from different modalities at the temporal level rather than using a frame-by-frame detection method.
ChapterThe Future of Autonomous Vehicle Technology
Episode#147 – Dmitri Dolgov: Waymo and the Future of Self-Driving Cars
PodcastLex Fridman Podcast
Object detection and classification is step zero in autonomous driving, but accurately predicting behavior and intentions of entities in the scene involves heavy machine learning and a hybrid approach including a combination of different types of ML models, model-free approaches, and physics-based systems.
1:52:46 - 1:57:05 (04:18)
Summary
Object detection and classification is step zero in autonomous driving, but accurately predicting behavior and intentions of entities in the scene involves heavy machine learning and a hybrid approach including a combination of different types of ML models, model-free approaches, and physics-based systems.
ChapterThe Future of Autonomous Vehicle Technology
Episode#147 – Dmitri Dolgov: Waymo and the Future of Self-Driving Cars
PodcastLex Fridman Podcast
The fundamental structure of behavior prediction, decision making and planning share a lot of similarities with language models, with a sequential nature, strong locality and larger context representation.
1:57:05 - 2:00:11 (03:05)
Summary
The fundamental structure of behavior prediction, decision making and planning share a lot of similarities with language models, with a sequential nature, strong locality and larger context representation. Waymo along with other Alphabet researchers have been pushing research in this space, with interesting breakthroughs in language models, supervised and unsupervised machine learning.