GM’s AI-Driven Tech Roadmap: Revolutionizing Vehicle Development

General Motors reveals its tech roadmap focusing on AI, advanced batteries, car computers, and autonomous driving.
Sterling Anderson doesn’t think humanoid robots should be building cars

GM’s Ambitious Leap into AI and Autonomous Driving

In a high-profile event in New York City, General Motors revealed its latest technological strategy. This plan emphasizes advancements in AI-based vehicle development, cutting-edge battery technology, powerful automotive computers, and autonomous driving capabilities.

While some aspects of this strategy were already public knowledge, such as GM’s work on hands-free, eyes-off driving automation and experimental new battery chemistries, the event’s primary aim was to demonstrate the integration of these elements. GM is introducing an “intelligence-defined vehicle” approach, leveraging AI, centralized computing, and virtual engineering to expedite vehicle development, reducing the typical four- to five-year timeline to just over two years.

This technological evolution comes amid significant industry challenges and narrow margins for mistakes. GM’s bold shift occurs as it competes with highly efficient Chinese EV manufacturers expanding their global footprint. Concurrently, waning EV demand in the US has prompted GM to reassess its electrification goals, balancing substantial investments in next-generation battery facilities with the ongoing expenses of software development.

GM’s commitment to unproven software architectures and autonomous technology represents a high-risk venture. If these intricate AI systems encounter difficulties or development delays arise, GM risks incurring substantial financial losses while lagging behind nimble competitors like Tesla, which have been software-centric from the beginning.

After the event, discussions with Sterling Anderson, GM’s chief product officer, delved into the company’s risky new ventures and strategies to guide this century-old Detroit powerhouse through an increasingly fast-paced, software-oriented industry.

What is the overarching theme of GM’s announcement?

“The car’s getting smarter,” Anderson stated succinctly. He elaborated on how vehicles are evolving into intelligent companions, citing improvements in the development process that allow for faster and more efficient design iterations. Anderson highlighted, “I kick off thousands of tests in the virtual world that give me feedback much quicker,” emphasizing the balance of speed and safety achieved through virtual testing.

How is GM utilizing AI models, and what drives the decision between internal and external data sources?

Anderson explained that the choice depends on the application. He noted that while GM has extensive in-house data, external data enhances model performance. The co-training of AI models with external data often leads to better outcomes, a practice GM employs extensively.

How is GM embedding AI models into vehicles, and what is the balance between onboard and cloud-based processing?

For safety-critical functions, processing occurs onboard to avoid latency and potential connectivity issues. Non-critical functions, such as OnStar AI interactions, may utilize cloud processing. GM is considering distilling these functions into onboard models to address latency, though this is not a safety requirement.

How has GM accelerated its digital transformation and the implementation of its “digital brain” concept?

Centralized computing and a common electrical system have been pivotal in speeding up GM’s software development. By standardizing the hardware set across its portfolio, GM enables software engineers to work more efficiently, akin to a company with a limited number of production models.

How does GM maintain brand distinctiveness while standardizing vehicle components?

Brand identity is preserved through aesthetics and interaction models, while underlying components are standardized for efficiency. This approach allows GM to provide a consistent ride quality across brands while differentiating on features, aesthetics, and interaction patterns.

What remains for GM to achieve fully autonomous driving, and how does it address complacency concerns?

GM is designing autonomous systems to be truly autonomous, avoiding terms like “eyes-off” that imply unrealistic levels of automation. The company favors a gradual approach, focusing on road-type-specific settings and expanding outward. GM ensures safety by not handing control back dynamically in critical situations, preferring to involve humans only when the vehicle is stationary and safe.

How does GM address privacy concerns with AI features that access personal data?

GM emphasizes designing privacy into its systems through end-to-end encryption and user control over data permissions. The company seeks to rebuild trust by simplifying authorization processes and ensuring that even internal errors would not compromise user data.

Is GM exploring factory automation and humanoid robots?

Rather than investing in humanoid robots, GM focuses on AI intelligence for specialized robotic tasks. The company partners with machinery suppliers like Fanuc while developing in-house solutions tailored to production needs.

What impact do GM’s technological advancements have on vehicle affordability?

New battery chemistries and a unified electrical architecture contribute to cost reductions, making GM’s fourth-generation EVs competitively priced. Virtual development processes further lower costs, benefiting both electric and internal combustion vehicles. As technology advances, GM aims to offer more capabilities without increasing costs, providing flexible subscription models for additional services.

Will GM’s sensor strategy evolve, or will it continue to use a multisensor suite?

GM plans to maintain its multisensor approach, combining cameras, radar, and lidar for optimal safety. This strategy offers advantages in diverse conditions, as relying solely on vision can be problematic in certain scenarios, such as dense fog.

Original Story at www.theverge.com