Introduction to AI in Smart City Industry – Episode 1

When we go to any smart city event, artificial intelligence is everywhere. It has been around since the 70s, but when computer vision became a commercial reality, AI became a very hot topic in smart cities. In the smart city industry, it is turning into a crucial tool, on occasions even a selling point, other times, it serves as deception from vendors who sell a simple algorithm as artificial intelligence to attract more buyers and score touchdowns in the industry, and sometimes, it is a worry for cities who want to protect their residents. There are many types of AI, but there are three we see the most in smart city solutions. Let’s explore them and start with the basics in this first video of an AI focused series. On one hand we have generative AI, it’s a new tech developed in recent years that can read text and create more text, audio, or video based on the initial input. ChatGPT is generative AI. It follows a structured set of data with a very defined context. And tries to mimic it every single time with a unique ouput while it’s remaining loyal to the source, which is the structured data and the context. Generative AI has proven very useful to vendors in the administration part of business, especially during a tender process. Public tenders are complex and generative AI can read and analyze texts from hundreds of proposals and related documents to help suppliers focus on the one they can win, drastically augmenting sales efficiency. It can also generate, under human supervision, tender answers, saving time. We’re also hearing of applications in water. On the other hand, there is also machine learning. The system is trained to identify certain aspects of the data and create alerts so tasks are performed following patterns. The machine learns from historical data and then analyses real time data to help in decision-making and even predict. It is very useful in predictive maintenance, as it can analyze and compare historical data to current one and alert of a possible malfunction. It can be used to gain accuracy in waste collection, for example optimizing routes based on accurate total weight prediction. We are able to detect things in the real world, we’re able to perceive things. And the way we do it is by looking at structured data and we identify it and we train models to do that. And then we have the star of AI in smart cities, the quarterback: Computer vision. It allows computers to analyze images and detect what it's been set to detect, flagging the anomaly. This aspect of artificial intelligence is the most used as it became commercially affordable 10 years ago thanks to cloud computing. It depends on our clients. We define with them specific scenarios. Imagine a client that wants to be alerted once a person falls. As soon as we detect in the stream, in the video stream such an event, we will send him an alert where he can see the person falling so he can take actions and react to the incident. It is used in cleanliness, mobility, safety… It can be used in smart cameras, or even with existing infrastructure, as all it needs to function in any vertical, is vision. The goal of AI in the smart city industry is efficiency, whether in traffic, ecological transformation, to save resources like water or energy, or improve safety and services. How does it do that? By taking repetitive tasks that would take a person a long time to do and giving it to the technology. So we have the ability today to have AI models, whether it’s a camera in a fixed corner of the street or in a vehicle, to turn it into value. What kind of value? We can understand traffic. We can detect anomalies. We can alert when we see an accident, people in distress, we can capture traffic violations. The AI is able to do it 100% of the time without getting tired at a 97-98% accuracy, way more than a human can and at a scale a human cannot. The ability to scale. What AI is giving the companies to do is to process that amount of information in an efficient manner. That is today one of the most critical things to do. Data processing before AI was taking 5 to 9 days, to give you an idea, just to go through the images, identify the defects and qualify the defects. Today, the same amount of data will take you 1 to 2 days. The lengthy task is given to the AI tool so the person can focus on other aspects of the project, so the use of the technology will allow both cities and vendors to work on other issues and save time, and time is money. Efficiency will increase, and more subjects will be solved. So to enable these high updates, with high frequency, same for the city. That’s really where our AI automation comes in, where you need to bring that intelligence to the edge to process immediately at the sensor what you see. And then really we have this opportunity of collecting only what’s relevant, bringing that back to the parties that update their products and services. In the smart city industy AI can enhance the capabilities of people. So rather than the technology substituting engineers, it would allow them to take on more projects a year, as some of the processes will be automatic and won’t require their supervision or efforts. Instead of having Bad Blood with AI, we could see the technology as an ally. I think we need to remember that AI is a tool, it should be a tool. I like to call it the copilot, at least for now. It’s in the service of improving our activities, our jobs. In the case of cities, we have a bit of challenges. Financing all these ideas is one. Talent is another, regulation. But they are mitigated by looking at, okay, we can follow the rules and regulation and we can find public private partnership programs to finance this project. We need to upskill people. So the traditional challenges that come with any innovation and new technology. We just got to go through the process of understanding and adapting and improving. We can't know the limitations of the technology as we don't know what's happening behind the doors of major company labs. When it comes to the dangers, the priority for citizens and cities is privacy as they know All Too Well, it can be compromised. For the Paris Olympics, the government will be testing computer vision without facial recognition, easing concerns. When talking about the tech, there is a trend among vendors to say that something is AI when it’s just an algorithm, so there’s a lack of transparency. This happens a lot in machine learning, because it’s a difficult and lengthy process to teach the computer. There’s unsupervised machine learning, but it lacks accuracy. Theoretically, everything can be done thanks to AI, but, that’s not the reality of it right now, there are limitations. Let’s ask ChatGPT. Manners cost nothing and better stay on AI’s good side in case there’s ever an ‘I, Robot’ situation. As we can see, there’s a lot more to explore regarding realities, limitations and policies of AI. We’ll do that in a second video, part of an AI series. So keep your eye on Kurrant.