Your car is a good listener. Maybe too good.
Just how much should a car know about its driver? As voice-enabled generative AI infiltrates the commercial auto fleet, there are questions about what to expect our cars to do for us, and what is being taken in exchange.
With industry giants like Qualcomm demoing systems that promise seamless integration between on-board voice assistants and real-time data pulled from the cloud, the car is fully transitioning from a mere transportation device to a computer on wheels – or, in the words of Jim McGregor, principal analyst for TIRIAS Research, “the ultimate intelligent autonomous platform.”
In a blog post this week, Qualcomm announced a partnership with SoundHound to develop and test SoundHound Chat AI for Automotive. Calling it “the first available voice assistant with generative AI capabilities,” Qualcomm says it will be added to the Snapdragon Digital Chassis concept vehicle and AI-based Snapdragon cockpit platform. The voice recognition and conversational AI capabilities of SoundHound’s end-to-end edge and cloud products make driver queries quick and seamless.
The platform draws on more than 100 sources of information, including third-party large language models, and selects the most relevant response among them, says the post. As illustrated by Qualcomm’s PR materials, a voice assistant might find a recipe, add the needed ingredients to a digital shopping cart, and have them ready for pickup at the driver’s local grocery store at a set time.
If the recipe is not right, AI can be trained to do better next time, training its algorithm through interactions to suggest healthier or more authentic options, and to generally cater to a user’s preferences.
But some observers say the trade-off is not worth it.
The data is coming from inside the car
Privacy watchdogs at Mozilla have released the results of their research into how automotive brands collect and use data and personal information. The assessment is blunt.
“Modern cars are a privacy nightmare,” says the report. What AI service providers think of as convenience-boosting mega-platforms that will change how people live, work, and transact, Mozilla’s privacy investigators call “data-gobbling machines” that have “unmatched power to watch, listen, and collect information about what you do and where you go in your car.”
The team researched 25 brands, including all of the largest and most popular automakers, and found that none of them passed the privacy test.
“All 25 car brands we researched earned our *Privacy Not Included warning label – making cars the official worst category of products for privacy that we have ever reviewed.”
Among the offenses were collecting too much personal information, sharing or selling data, and failing to give drivers adequate – or any – control of their data. The researchers were unable to confirm that any of the automakers met their Minimum Safety Standards. And a few brands earned extra scorn.
“Tesla is only the second product we have ever reviewed to receive all of our privacy ‘dings’,” said the report, pointing to a number of deaths and crashes that have been blamed on the electric car company’s AI-powered autopilot.
As the Mozilla team notes in its conclusion, driving is a necessity for many people, “so unlike a smart faucet or voice assistant, you don’t have the same freedom to opt out of the whole thing.” A 2019 survey by the American Auto Association found that the average American spends nearly an hour a day in their car – plenty of time to get to know someone very well, indeed.