OpenAI GPT-6 Astra Drives Toyota Corolla Through In-N-Out Drive-Thru
Three artificial intelligence engineers navigated a 2024 Toyota Corolla through an In-N-Out Burger drive-thru in the San Francisco Bay Area using a text-based AI model. Aditya Ramabadran, Simon Mahns, and Tobias Gessler from the startup Axiom connected OpenAI’s GPT-6 Astra chat interface to a server linked to windscreen-mounted cameras and the car’s power steering system. A safety driver kept one foot poised above the brake pedal during the experiment. The model, typically used for generating text and code, guided the two-ton vehicle to the take-out window without any prior physical training.
Startup Focus Shifts to Physical Reasoning and Visual AI
Traditional autonomous vehicles rely on algorithms specifically engineered and trained solely for driving tasks. The Axiom experiment demonstrates that general-purpose language models are beginning to develop a rudimentary understanding of the physical world. While current AI systems excel at answering complex questions online, they typically struggle when placed in physical environments. Several researchers have launched startups to tackle this gap in physical reasoning. Andrew Dai, CEO of Elorian AI and former Google DeepMind researcher, states that enhanced visual reasoning will enable new applications such as home robotics and systems that evaluate customer experiences in restaurants.
Frequently Asked Questions About AI Drive-Thru Navigation
How did the AI control the Toyota Corolla?
The engineers connected a chat interface running OpenAI’s model to an external server. That server linked directly to windscreen-mounted cameras and the vehicle’s power steering system to translate the model’s outputs into steering actions.
Who conducted the In-N-Out experiment?
AI engineers Aditya Ramabadran, Simon Mahns, and Tobias Gessler from the startup Axiom carried out the test in the Bay Area using a 2024 Toyota Corolla.
Keep reading