Researchers at Ford are testing a hybrid gas-electric car that makes an educated guess at where you’re going whenever you turn the key.
They installed software that draws on prediction technology developed by Google in a plug-in hybrid Ford Escape SUV. To make the car more energy-efficient that software directs the car’s computer to tweak how its electric motor draws power from the vehicle’s battery and gas generator during a drive according to the trip a driver is expected to make.
“The system keeps track of how a person uses their car and builds a predictive model in the cloud, using Google’s prediction technology,” says Ryan McGee, of Ford’s Dearborn, Michigan, research labs. “When you start the car, it asks that model, ‘Where are we going next?’”
Ford’s prototype makes use of a Google service called the Prediction API to create, store, and query that model. When data is uploaded to the service, machine-learning algorithms build a model that can be used to predict future additions to the data set.
In the case of the Ford prototype, the car uses a wireless Internet connection to supply the prediction service with the vehicle’s current location and the time. It receives back a ranked list of likely trips. Based on that list, the software can inform the car to change the way its engine management software juggles gas and electric power consumption over the trip. “It might use electric energy earlier in the trip, or save it for the end,” based on rules set by the driver, or derived by the car’s software from past experience, says McGee.
If a driver’s commute to work contains a section near the end with a lot of stop-start urban driving, the car might decide to avoid using its battery early on in order to handle that section most efficiently later. The car could conceivably tap into advice or rules from local governments, says McGee, and juggle its power sources so as not to use gas in pollution control zones.
Google launched its prediction service last year, but this week the company will open it up to anyone, and even guarantee its reliability for a fee to encourage its use in real products. Google’s engineers have also made the system capable of refining an existing model on the fly, says Travis Green, product manager for the service. “You can then stream in training data,” says Green. “That means as a person drives around to new places, you can add that in to the predictive model of their habits.”
McGee and colleagues are currently driving the modified Escape around Ford’s Dearborn campus, testing, among other things, how quickly Google’s prediction service can learn a person’s habits. “It would probably learn my daily commute very quickly,” says McGee. “But patterns like that I play soccer once a week in the winter is more complex.”
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
Data analytics reveal real business value
Sophisticated analytics tools mine insights from data, optimizing operational processes across the enterprise.
How to fix the internet
If we want online discourse to improve, we need to move beyond the big platforms.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.