Robotaxi Lobbying Wars and Advances in Robot Foundation Models

로보틱스 | Sat Aug 22 2026 00:00:00 GMT+0000 (Coordinated Universal Time) | 2 sources

Waymo and Uber intensified autonomous driving regulatory lobbying, while Generalist AI demonstrated a rapid-learning robot model.

Analysis

[Waymo] doubled lobbying spending to expand robotaxi services ^[1]

[Uber] pursued lobbying for a hybrid network approach ^[1]

Attempting to recover after shutting down its own autonomous driving business in 2020
Invested over $10 billion in equity investments and robotaxi agreements over the past year
Advocated for a model combining human drivers and autonomous vehicles
Proposed allocating at least 85% to human drivers in the New Jersey pilot program

[Generalist AI] unveiled a robot that learns from demonstration videos without task-specific training ^[2]

Performs various tasks after watching short demonstration videos without separate training
Recorded an average success rate of approximately 59% on demonstrated tasks
CEO compared it to prompting GPT-3 with new tasks
Data collection strategy for large-scale physical interactions not tied to a specific robot

[Generalist AI Robot Demonstration] showcased improvisational problem-solving capabilities ^[2]

Improvised behavior of using a dustpan to knock blocks into a bowl when a broom was missing
Withdrew bills from another wallet after watching a demonstration video of opening a wallet zipper
Switched the gripper from right hand to left hand when grasping failed
Attempted to sweep objects using a banana placed in front of it

[Generalist AI Team and Data Strategy] built an independent training pipeline with founders from DeepMind and Boston Dynamics ^[2]

Co-founders Pete Florence (CEO)
Andrew Barry (CTO)
and Andy Zeng (Chief Scientist)
All three founders have experience at Google DeepMind and Boston Dynamics
Collected demonstration data directly from humans using camera-equipped gripper gloves
Built AI models from scratch without relying on open-source language models