CES 2020: Location-Based AI is Enabling an Efficient Future

Location-based data is key to many of the efficiencies promised in smart, AI-enabled cities. HERE Technologies got its start in location data in 1985 when, as Navteq and later Nokia, its goal was to digitize mapping and pioneer in-car navigation. In 2015, HERE was sold to a consortium of German automakers and currently has nine direct and indirect shareholders. The company now creates 3D maps and other location-based solutions. During CES, HERE senior VP development & CTO Giovanni Lanfranchi described how the company ran a hackathon in Istanbul that challenged ordinary citizens to come up with new location-based solutions.

Lanfranchi and HERE Technologies director of data science Michael Kopp reported the tremendous success of engaging local residents to help solve problems that they knew intimately. “We mapped traffic flow [of Istanbul] and used a lot of sensors,” said Lanfranchi. “Neural networks managed to capture the component that was temporal and was very good sequentially to fill in the gaps.”

Kopp noted that many entries used a similar neural network approach, but stated that one of the surprises was seeing “all the other methods you might think didn’t work.” “You need a lot of trial and error,” he said.

Axios chief technology correspondent Ina Fried, who moderated the conversation, noted that not all AI is equal. “Yes, you can use AI to predict traffic — poorly,” she said. “We’re in a hype stage where there is important groundwork and AI is used effectively. But it’s not necessarily better because it’s AI.”

Lanfranchi stressed the importance of the consumer experience, whether the solution is AI-enabled or not. “We believe in an open community around AI,” he said.

Fried challenged them to predict what a city will look like in five to 10 years. “Not the one Toyota is building from the ground up,” she said. “Toyota is building a city of the future near Mount Fuji. And that’s fascinating. But no other city, whether it’s Berlin or São Paulo, will work like that because of their existing roads, infrastructure.”

Kopp answered that, although “this is very crystal ball,” he looks at the successes that AI has had in generative (GANs) methods. “Maybe we don’t have buses with fixed stops,” he surmised. “You can control it more because it’s predictive. The aim is to have a much more sustainable city — to bring people from Point A to B, but with less carbon consumption.”

Lanfranchi added the three characteristics he considers about AI in a future smart city vis-à-vis location. “First, a society where all the data is private by design,” he said. “If I provide data, I know what the purpose is. Second, it’s very decentralized and personalized. Third, it provides transparency for the entire community.”

“I want that world,” said Fried. “But, can I have that with privacy and transparency?” Lanfranchi noted that privacy is “not a binary qualifier. “It’s a spectrum of decisions,” he explained.

The three discussed how difficult it is to actually make data anonymous. “With four places a vehicle has gone, you know who it is, in 80 percent of the cases,” said Kopp. “At the moment, no one has the answer [to provide efficiencies and privacy]. AI is a powerful tool that can be used to de-identify data, but then blockchain can prevent it. It’s an arms race going on.”