large language model, and you train it and refine it on a specific domain,” Richardson said. “We’ll take a base model and train it on the vehicle’s specifications, distill that down, and have that running on the vehicle.”
While GM has a close relationship with Google and will already be implementing Gemini into certain vehicles, Richardson said GM plans to test several foundational models from other AI firms, which could include OpenAI, Anthropic, and others.
Richardson said drivers will be able to control what information the assistant can access and use, and it can learn from your habits to offer personalized recommendations. GM’s emphasis on user controls is notable given the company’s recent controversy over selling customer driving and geolocation data to insurance brokers.
Richardson said any data GM gets from drivers goes directly towards improving the product and won’t be sold to bring in additional revenue for the automaker. Over the last nearly two years, GM has brought on a new data team — including Christina Montgomery, who spent 30 years as IBM’s chief privacy and trust officer — to put standard processes and data governance technology in place.
“Everything that we’re going to do is going to be driven by customer consent, so you can always opt in or opt out,” he said. “Our viewpoint is that data and privacy has to be built into everything that we do.”
This article has been updated with comments from Dave Richardson.
