Correction: A previous version of this article erred in describing Andrei Khurshudov’s role at Yale University. He is a faculty member. We also have corrected the subhead, which misstated his views regarding data privacy related to smart cities' data collection, and we have clarified that Khurshudov’s comments do not represent his employer.
U.S. cities currently collect hundreds of thousands of data points a day, from license plate information to pedestrian volume to traffic flow.
In the last decade, the sensors used to collect this data and the storage to house it have become so affordable that they’re ubiquitous. Because of this, much of the collected data goes underutilized, says Andrei Khurshudov, a faculty member at Yale University’s School of Engineering and Applied Science and director of IoT analytics and artificial intelligence at Caterpillar. (Khurshudov’s comments do not represent his employer.)
What makes a city smart, Khurshudov says, is the use of information communication technology and the Internet of Things to become more sustainable, efficient and able to improve the quality of its services. Smart Cities Dive spoke to Khurshudov about how cities can better utilize their data, what privacy concerns might exist and what role AI is playing in developing smart cities.
Editor’s note: This interview has been edited for clarity and length.
SMART CITIES DIVE: Many cities collect massive amounts of data through technology, but what turns that information into smart cities strategies?
KHURSHUDOV: The reason for overcollection is that if we don't store data today, we'll have to start from scratch. Sometimes we know what we want and we have the data, but we don't know how to achieve that goal because our algorithms or techniques or computing power or something else is not available.
The list of objectives for smart city projects is roughly the same for all cities around the world: improving the quality of life, economic competitiveness, operational efficiency and sustainability of the city and the people living in it. But cities prioritize these objectives differently because none of them have unlimited assets or funds.
Do you have privacy concerns around cities collecting data and incorporating artificial intelligence?
If prompted, people always say yes, we are concerned about privacy, but in reality, we feel fairly comfortable with closed-circuit television video surveillance and data collection. It does not concern me. We need to recognize that smart cities are built on the foundation of data collection. If we want smart cities, we have to give them data about you and I.
AI is just another way to compute things and figure out the relationships, and so on, but there is one difference: Most AI algorithms are opaque — they’re what’s called black box algorithms — because the question goes in, the data comes in and the answer comes out, but you do not see how this answer was derived. So when you put AI algorithms in charge of decision-making, especially for critical decision-making cases, that is something to keep an eye on. There's always some unpredictability present in AI algorithms, but sufficient testing and validation should take care of it.
How will increasing AI use impact how smart cities operate?
Well, think about license plate recognition: It's actually relying on machine learning algorithms to recognize those plates, and face recognition is just image recognition done by machine learning techniques. It's rare for cities to have their own large data science teams to build their own analytical solutions, so they typically acquire them in the smart products they purchase. The smart lights and sensors come with their own software. AI could be used for decision-making in the presence of a lot of data. So it's definitely a big part of the analytics software that’s behind the dashboards city management uses to make decisions about their city. Data collection and AI start on the street and end in an office space where humans still need to make the decisions.
