LocationsInterview with
Muhammad Ahsan Younas
Deputy Inspector General of Police & Managing Director, Punjab Safe City Authority
"We have launched the world’s first virtual women police station using AI-powered voice recognition system that connects women directly to the station.”
How has the Punjab Safe City Authority benefited from applying AI in promoting urban resilience and development? What have been the key successes so far?
It is important to define urban resilience. For us, it refers to the city's ability to anticipate, respond and recover from emergencies, crimes, and public safety threats while maintaining essential services and infrastructure. In the last year we have experimented with AI across several urban resilience initiatives with successful results.
We have launched the world’s first virtual women police station which aims to make safety services more accessible. We connected the virtual police station with the emergency response system that is accessed by dialling ‘15’. Initially, we had used an Interactive Voice Response (IVR) system that allowed women to connect with virtual women police stations directly, however, there was a lot of traffic of male voices on the station. Therefore, we adopted the first AI solution to conduct a gender voice check.
We created two options in our IVR system, option one leads to the emergency option and the second connects to the virtual station. When choosing option two, the machine will ask for your name and city. Once an individual answers, within the first three to four words, the AI determines whether it is a male or female voice, and they are directed to the correct options. The male voice is referred to the emergency option, while the female voice proceeds with registering her concern. As a result of the AI-enabled automated gender voice recognition, the virtual women police station remains accessible exclusively for women, enhancing safety and reducing misuse.
We have recently tested a Characteristic Based Object Retrieval (CBOR) system. This system is designed to enhance public safety through real-time person or object identification and tracking. The CBOR uses advanced machine learning algorithms and computer vision technology to identify number plates or facial characteristics that are unrecognisable by traditional surveillance systems. We can conduct prompt-based searches by applying specific prompts such as, ‘find male wearing black clothes and red helmet’, or even something like, ‘find black Toyota Corolla with ‘MashAllah’ written on the back’.
We have also developed an AI-powered e-challan systems which automates the detection of traffic violations and issuing e-tickets. The system can identify 20 different types of traffic violations, this is the largest in the region.
We are also working towards issuing alerts for public mob and street fight detection. We are using a pitch and sentiment analysis on our ‘15’ emergency calls. In a scenario where our agents or any of the citizen use abusive language, the system can detect it and generates a report. We also have a Chatbot that extracts this information.
Chief Minister Punjab Maryam Nawaz Sharif visited the Punjab Safe City Authority headquarters.
What has success looked like for Punjab Safe City Authority in terms of accuracy of these AI-led initiative?
The accuracy of AI-led initiatives is around 90 to 93 and 94 percent. At the virtual station after 100 calls, one male call may land.
You may have seen on the media that our AI issued an e-ticket for not wearing a helmet to a vehicle. The camera was actually focusing on the driver’s right where there was a motorcyclist who was without the helmet, the system confused the two.
However, the best thing about machine learning is that it improves in time. We worry about these glitches because we have a sensitive job to perform, but we also ensure to correct them immediately.
I believe, our human led system should be so dynamic that it addresses any AI false alert immediately and correctly. The moment we get to know through our own surveillance system or the person raising the complaint of a false alert we must have a dynamic response system in place to correct the mistake. We have a strong system for any grievance against e-ticketing and immediately address the issues and withdraw false tickets issued. We take a lot of pride in the fact that we immediately accept mistakes and course correct.
We have spoken in detail about the initiatives that are being implemented currently. However, looking towards the future, how do you envision the Punjab Safe City Authority leveraging AI to make cities smarter and more resilient?
We are in the process of launching responsive drones on our ‘15’ emergency systems. In any critical situation we will have a drone squad which can be air borne under 90 seconds. This initiative will backstop human effort with precision and agility.
Drones can be deployed for oversight in situations that necessitate an immediate safety and security response. They can be present at the situation and relay real-time information back to the center. We aim to incorporate within the drone a telecommunication system that allows making public announcements.
We are also looking into the streetlight issues. In places where we have our cameras installed, these cameras should detect whether a streetlight is working. It is a very small solution which we will be giving to the government soon. Similarly, we are working on a project where our camera system will issue an alert if it detects students outside during school hours.
There are 101 things you can imagine with AI and the cameras, we do think that apart from security and dealing with crime, we want to contribute positively to society and implement an inclusive government approach. That is why we have given access of our surveillance systems to all the Deputy Commissioners, helping them see the civic services being deployed and the impact on well-being in society.
How do you perceive the public's readiness to use and access these AI- based services aimed at improving their quality of life? Are they ready?
It depends upon the efficacy and effectiveness of the service delivery. A good service will allow people to be open and accepting of AI-led services. If the accuracy of e-ticketing does not improve and there are false alerts, then our credibility will be low. Similarly, if the Chatbots are working well and answering public concerns, then there will be more readiness to accept AI-led services. Therefore, I think it depends on our ability to put credible and transparent AI system before the public which will determine their readiness. Generally, I do not find many complaints regarding AI-led systems. It is true that people do not like to be surveilled and taken to accountability. But it is necessary.
We are also trying to ensure our algorithms are not biased and are transparent. We do have to rely on previous data which is inherently biased and that seeps into our data systems. We are aiming to eradicate this bias and develop inclusive and bias-free algorithms.
© PSCA
Has there been any public awareness, campaigns or engagement efforts to build trust in AI-powered solutions among the public?
Yes, there are many. The virtual station has helped over 340,000 women in eight months. This shows readiness as the public has accepted AI-powered solutions. And similarly, we have issued e-tickets amounting to more than PKR 1 billion. People are complying with AI-generated tickets by paying their fines. This means that the public is receptive and accepting of AI. I always say that if things work correctly, it will determine its acceptance. The more transparent and efficient the service is, the more accepting people will be. I do not think Pakistanis as a nation are aversive to technology. They are apprehensive but not averse.
Do you anticipate any challenges associated with an AI urban management?
Yes. We have multiple challenges; the biggest one is scrutiny. A false situation can land us into trouble and embarrassments, which can further lead to legal cases and public outcry. With technology, our approach is to be humble with the citizens.
And from a technological point of view, what is the biggest challenge in integrating AI within the existing systems? Be it from a point of view of infrastructure or human resource?
I think the challenge is not in creating new systems or integration within traditional systems, the real challenge is the mindsets. We must overcome the mindset which hinders some of the integrations. There are instances when people cast a bad light on good initiatives due to personal agendas, or perhaps people who might not get their desired or expected outcome from the system can present it as a biased system.
I do not think there are technology issues. I think the issue is of the mindset, but the silver lining is that with each passing day we have more successes.