AI has been impacting every aspect of our lives, particularly now it has become generative, and it’s going to affect the automotive industry dramatically too. Chinese manufacturer XPENG even reckons there is an AI-defined future ahead for vehicles, and Volvo is using an AI technique called “gaussian splatting” to help train its safety models. Some of these changes won’t be in the car itself, but how its maintenance and road use are managed – particularly in fleets. I talked to leading vehicle telematics company Geotab’s Senior Vice President Edward Kulperger about how AI is improving reliability and making roads safer.
Geotab’s Long History With AI
“We've been delving into AI for over a decade,” says Kulperger, arguing that Geotab was ahead of the curve. “Neil Cawse founded the organization from his basement 25 odd years ago. Now we're 2,500 people and serving some of the largest companies in the world. He's been investing in things like data programs and scientists since 2013, including AI. He also invested in electric vehicle technology, acquiring a company called FleetCarma that does EV tech in 2018.”
“They were working with different fleets to help them transition to electric vehicles ahead of the time,” says Kulperger. “Now we’ve developed some fascinating electric vehicle suitability assessment tools as well as infrastructure assessment tools.” These enable, for example, a large last-mile delivery company to assess what it might need to move from internal combustion to electric. The tools then support the transition and how the company can ensure sufficient power for operation. The company has also provided some of the most comprehensive EV battery longevity data, which has shown that electric cars are likely to last much longer than some people assume.
AI is taking an increasingly important role in Geotab’s tech-focused investment strategy. “The godfather of AI, Geoffrey Hinton, spoke at Geotab’s annual conference, Geotab Connect, in 2018,” says Kulperger. “He has just won the Nobel Prize in physics, but he's a University of Toronto professor. We're based just outside of Toronto and bridged him into our conference, where he did one of the keynotes. We've evolved the organization from data to EV to providing a platform to a wide ecosystem. We are now leveraging AI tools to extract information from our platform using large language modeling.”
Geotab has been using AI technology for a few years now, but while LLMs have sent shockwaves through creative circles, they still have a problem with hallucination, where Generative AI invents information. That’s an issue if you’re using a search engine but it could be very serious if it delivers incorrect information to an automotive management system. “Obviously that’s incredibly tricky,” says Kulperger. “But we have 150 data scientists looking at the data.” These data scientists verify that output is correct. “We do a full monitoring program within our data.”
Predicting Breakdowns With Geotab AI
“We’ve developed other AI tools to look at last mile organizations doing deliveries in our communities,” says Kulperger. “They're starting and stopping their vehicles thousands of times a week. That puts an incredible amount of stress on your battery, your starter and your alternator. We’ve looked at patterns of this, such as the time it takes to start a vehicle from that electronic pulse to hit ignition to starter to alternator. Through that sequence of events that's in millisecond detail, you can understand when things are going to break down. Some of these large last mile companies will pull vehicles off the road and change either the starter, alternator, or battery before they break down, saving millions of dollars for them. If a vehicle breaks down on the road, going out and servicing that is a big problem. We developed that about four years ago now. We have taken these learnings and pushed them down into our products so that fleets of all sizes can leverage them.”
This is just one example of how Geotab has been harnessing the power of AI, and Kulperger reckons the pace of development will speed up. He sees public and private sectors increasingly pumping data, processing power and energy into AI. “Things are exponentially accelerating beyond even what we thought,” he says. “Predicting the next year is going to be incredibly challenging.” Geotab has developed a platform to facilitate its clients harnessing its data, including with AI. This platform can be used for innovative safety tools, vision telematics and AI models built on top of Geotab’s software.
This approach is paying dividends when analyzing accidents to ensure they can be mitigated as cost effectively as possible. “We have our own 150 data scientists that are focused on extracting data rapidly to process it for partners to be able look at what happened in an accident and then using neural networking to dive into our data lake to see if we’ve seen similar accidents in the past,” says Kulperger. “They can see what happened in those accidents around speed, G-force, even things like RPM, braking, time of day and road conditions. They can then reconstruct that accident again for customers to help them save money in terms of the repair. We'll understand deeply what happened with that accident. Because typically we'll have seen that already in our data lake.”
Geotab has been harnessing Generative AI to make its data more accessible to clients. “We built the foundation for that and enable our partners to create simple queries such as asking how much energy they used last week in Birmingham or London with their vehicles,” says Kulperger. “That query will go out to the model and come back with an answer. We also provide how we got that answer. That’s important, but then when you look at our data lake beyond that, we have around 80 billion data points coming into our ecosystem every day. That’s an astronomical amount of data.”
There’s going to be a lot more to come from applying AI to Geotab’s huge dataset. “We have methodologies for partners, obviously privacy and security are paramount,” says Kulperger. “But then we enable partners to leverage our data lake, aggregated and anonymized, so organizations in the public world can improve their infrastructure. For example, they can know where to put speed limit signs and curtail accidents that are happening and dangerous hotspots. We're only hitting the tip of the iceberg of uses today.”