Toyota Research Institute’s Vision for Seamless AI-Human Integration

Putting Humans First

Gill Pratt possesses a unique ability to simplify intricate concepts, making him an ideal spokesperson for Toyota Motor Co. as its chief scientist and CEO of the Toyota Research Institute (TRI).

In a recent gathering at TRI’s Silicon Valley headquarters, Pratt engaged with tech reporters and analysts, including SAE Media, to discuss the increasing attention paid to artificial intelligence (AI). It’s a key area of research for TRI’s 200 scientists and engineers.

With humor, Pratt addressed concerns surrounding ChatGPT, the controversial chatbot developed by OpenAI in late 2022, and its potential to write term papers for college students. He observed society has become less naive and more skeptical of technology, particularly due to the initial hype around self-driving vehicles – a hype that TRI did not participate in.

Pratt recalled TRI’s first CES presentation seven or eight years ago, during which it emphasized the challenges of autonomous driving and the uncertainty surrounding the timeline for achieving SAE Level 4 automation. The candid approach has defined TRI since its inception in 2015, backed by a $1 billion investment from Toyota.

As a small portion of Toyota’s roughly $10 billion annual R&D spending, TRI focuses on machine learning, robotics, energy and materials research. Its human-centered AI tackles issues like climate change, an aging population and human understanding, ultimately aiming to facilitate individual happiness and societal harmony.

The directors of TRI’s primary divisions believe in the jidoka principle – “automation with a human touch” – which has guided Toyota since its founding in 1926. Pratt emphasized that technology should enhance human work rather than replace it, maintaining the dignity of human labor. Accordingly, TRI’s mission is to undertake high-risk, high-reward research projects.

Acknowledging the likelihood of failure in such projects, Pratt argued a failure rate of around one-third indicates they are pursuing appropriately challenging endeavors. Outgoing Toyota CEO Akio Toyoda tasked Pratt with a simple directive: “Surprise me.”

This expectation persists under incoming CEO Koji Sato, as TRI continues to take risks and push boundaries, maintaining Toyota’s position at the forefront of innovation and IP generation.

AI Gaining Traction in Vehicle Development and HMI Design

In the vehicle development community, artificial intelligence (AI) is increasingly being recognized as essential for vehicle and systems development. By assisting humans in solving intricate engineering challenges, AI offers reduced testing and validation time, accelerated time-to-market, and lower costs.

AI also plays a crucial role in machine learning and the design of human-machine interfaces (HMIs) in vehicles and their functionality.

Adrien Galdon, director of Machine Learning (ML) at Toyota Research Institute (TRI) and an adjunct professor of computer science at Stanford University, stated the history of AI has been marked by hype. However, Galdon is optimistic about the growing usefulness of web-based tools, particularly when they involve human oversight and promote collaboration between humans and machines.

Galdon highlighted the importance of human involvement in AI development and applications. He explained the concept of Foundation models, a new generation of ML models that are trained on diverse and general datasets.

These models can adapt to individual users through continuous interaction and in-context learning. This approach represents a shift in AI from domain-specific systems to broader AI that can work across domains and problems.

Max Bajracharya, senior VP of Robotics at TRI, agreed machine learning should be about generalization and user adaptation. Bajracharya’s 50-person robotics team is focused on addressing the challenges of an aging society, with Japan’s population of people aged 65 or older expected to reach 50 percent by 2050.

Charlene Wu, TRI’s director of Human-Centered AI, emphasized how technology is used is as important as the technology itself. She discussed an example from her team’s work in human-centered AI and future product innovation.

By leveraging generative-AI tools, designers can explore numerous visualizations of a single input, allowing for increased creativity and efficiency. This approach exemplifies how humans and AI can interact productively.

Human Interactive Driving: Enhancing Driver Experience with AI

Avinash Balachandran, director of the Human Interactive Driving (HID) division at the Toyota Research Institute (TRI), seeks to create an exceptional driving experience for people by incorporating AI.

During a visit to TRI’s engineering garage, he showcased three development platforms built by Toyota Racing Developments (TRD): a Lexus LC equipped with data-acquisition gear, an open tube-framed electric mule vehicle with wheel-hub motors and by-wire controls, and a race-kitted Supra designed for autonomous drifting and data acquisition. Above the garage is a multi-axis driving simulator and control room, co-developed with TRD.

Balachandran’s team aims to establish a relationship between AI and drivers that mirrors the enjoyment and improvement derived from other activities like skiing or biking. Before joining TRI, Balachandran worked on Uber’s early self-driving vehicle program and contributed to Uber’s first autonomous service in Pittsburgh in 2016. His division’s focus is on the intersection of humans and vehicles, starting with understanding human behavior.

In the short term, Balachandran aims to make advanced driver-assistance systems (ADAS) in the SAE L2/L3 space seamlessly reassuring and intuitive, without noticeable technological interference. This approach aligns with Toyota CEO Akio Toyoda’s principle that driving should be enjoyable.

Balachandran emphasized that fun-to-drive is crucial for achieving happiness and mastery in driving skills. While Toyota is developing autonomous vehicle (AV) solutions, the current focus is on advanced driver assistance technologies.

TRI co-founder Gill Pratt envisions technology that adapts to each individual as a key outcome of TRI’s work. He stresses the importance of humility for technologists and cautions against assuming they have all the answers. According to Pratt, it is crucial not to impose decisions on society just because of technological expertise.