General Motors Sets Ambitious Course with AI-Powered Vehicle Innovations

Last week, General Motors hosted a high-profile event in New York City, unveiling its advanced technology roadmap focused on the integration of artificial intelligence in vehicle design, next-generation battery technology, powerful onboard computing, and autonomous driving capabilities.

While some details were already public knowledge, such as GM’s endeavors in hands-free driving automation and new battery prototypes, the event primarily aimed to demonstrate how these various components interconnect. With its new “intelligence-defined vehicle” approach, GM is targeting a drastic reduction in vehicle development timelines, hoping to trim the traditional four-to-five-year process to just over two years.

However, this technological transformation comes at a challenging time for the automotive industry, characterized by fierce competition and shrinking profit margins. GM is navigating an influx of highly efficient Chinese electric vehicle manufacturers gaining traction globally, all while facing reduced EV demand in the U.S. This has led the company to reassess its electrification strategy, balancing significant investments in next-gen battery facilities with ongoing software development costs.

Ultimately, GM’s commitment to untested software systems and autonomous technology presents a risky gamble. If these intricate AI frameworks fail or encounter delays, GM could deplete resources quickly, potentially lagging behind more agile competitors like Tesla, who have been software-centric since their inception.

Following the event, I had the opportunity to speak with Sterling Anderson, GM’s Chief Product Officer, regarding the company’s new ventures and how he envisions guiding this historic automaker through an increasingly rapid, software-driven landscape.

Insights from Sterling Anderson

This interview has been lightly edited for clarity and length.

Unifying Vision

Interviewer: Your announcements are quite broad. Is there an overarching theme?

Sterling Anderson: The car is getting smarter. It’s becoming a reliable, intelligent companion in various ways. I’m particularly excited about the faster and improved development processes our team is implementing.

AI in Vehicle Design

Interviewer: What AI models are you employing?

Sterling Anderson: It depends on the application. We utilize transformer-based foundational models for generative design tasks using our extensive in-house data, while also enhancing them with external datasets.

Embedding AI Models

Interviewer: How are these models integrated into vehicles? What balance exists between localized and cloud operations?

Sterling Anderson: Safety-critical functions require onboard processing to avoid latency issues. Non-critical tasks may use cloud processing, but we aim to improve latency and consider moving some models onboard for better user experience.

Standardization Across Brands

Interviewer: Given GM’s diverse brand portfolio, how do you determine what should be standardized across models?

Sterling Anderson: We aim to commonize components while preserving brand identity through aesthetics and interaction models. This ensures optimal engineering without sacrificing individual brand characteristics.

Challenges in Autonomy

Interviewer: What hurdles remain before achieving full autonomy?

Sterling Anderson: Our approach involves designing systems that account for situations where the driver may not be completely attentive. We distinguish between fully autonomous and driver-assist features carefully.

Data Privacy Considerations

Interviewer: How do you ensure customer trust regarding data privacy?

Sterling Anderson: We prioritize privacy by using strong encryption and allowing customers control over their data. Our architecture prevents unauthorized access, even by staff.

Cost-Effectiveness and Automation

Interviewer: How will these innovations lead to more affordable vehicle options?

Sterling Anderson: Advancements like LMR (lithium manganese rich) battery chemistry allow us to offer better pricing without relying on subsidies. Additionally, standardization reduces manufacturing complexity and costs.

Future of Sensor Technology

Interviewer: Will GM continue using a multi-sensor approach, or consider a vision-only strategy?

Sterling Anderson: We believe maintaining a multi-sensor suite is crucial for addressing various driving conditions effectively. Each sensor plays a vital role in enhancing safety.