Solutions

AI for Smart Cities

In the past 70 years, AI has been a rapidly growing field and is being applied nowadays in a wide range of applications. Artificial intelligence (AI) plays a crucial role in enhancing smart city initiatives, offering numerous benefits such as improved water distribution, efficient energy management, better waste handling, and alleviation of traffic jams, noise, and pollution. The primary efforts in smart city development have centred on data generation and the acquisition of new insights into the intricate and dynamic nature of urban environments. (H.M.K.K.M.B. & Mittal, 2022). Since 2008, cities have been incorporating artificial intelligence (AI) to enhance decision-making processes and contribute towards the achievement of the Sustainable Development Goals (SDGs). (Ingwersen & Serrano-López, 2018) What is AI? Artificial Intelligence (AI) refers to the development of computer systems that can perform tasks that typically require human intelligence, such as visual perception, speech recognition and natural language processing. AI technologies involve various techniques, including machine learning, deep learning, and natural language processing, to enable machines to learn and perform tasks without explicit programming. Essentially, artificial intelligence (AI) is a technology designed to generate results aimed at achieving a specific goal. This goal is a human-devised aim transformed into a mathematical format. The outputs generated by AI can include forecasts, suggestions, or choices. (OECD, 2023). The concepts of algorithm, AI system, AI ecosystem, and AI represent varying levels of scale and complexity. An algorithm stands at the most detailed level, acting as a set of instructions that transforms input into output. An AI system refers to a singular application that processes specific input to generate its unique output. An AI ecosystem describes a complex network of interconnected AI systems that communicate and interact with one another. Lastly, AI is the most general term, encompassing the entire spectrum of technologies, methodologies, and systems within the field. (Leal, et al., 2022) There are two main categories of AI systems: symbolic and statistical. Symbolic AI operates based on explicit rules and logic to reach conclusions, while statistical AI, in contrast, learns from patterns in data, using induction rather than deduction from set rules. Each method has its strengths, depending on the application. However, the recent surge in AI's capabilities and its widespread adoption can be attributed mostly to a subset of statistical AI called machine learning (ML). (Climate Change AI, 2021) AI for Smart Cities The potential benefits of AI for city planning and management are manifold. AI offers new tools and methods for analysing, modelling and simulating complex urban systems, thereby having the potential to create more sustainable, resilient and liveable smart cities for their citizens. By monitoring and analysing data, AI systems can provide recommendations to improve resource usage, help reduce congestion and improve traffic flow, detect potential security threats, reduce energy consumption, promote waste reduction, etc. AI-powered systems can also provide personalized services to citizens such as personalized recommendations for restaurants, events, and activities based on their preferences. However, it is important to ensure that AI is implemented responsibly, with a focus on transparency, accountability and privacy. Such technology can still provide accurate analytics, while remaining non-intrusive. Implementing AI technologies could be crucial in overcoming worldwide social, economic, and environmental challenges. Although each city has its unique characteristics, urban areas are pivotal in driving societal changes, especially in terms of digital advancements. Cities are hubs for the aggregation of people, employment, research, wealth, and recreational activities. They offer increased access to opportunities for a larger segment of the population while also concentrating both societal problems and environmental concerns. Given their role in international networks, the advantages and disadvantages of adopting AI technologies have implications that reach far beyond their geographical limits. (Leal, et al., 2022) As urban areas face pressing issues related to resource allocation, governance complexities, socio-economic disparities, and environmental hazards, innovation becomes essential in addressing these evolving challenges. (Yigitcanlar, et al., 2021). To maximize AI's potential for urban improvement, municipal governments need to establish conditions that support sustainable and inclusive growth. AI governance aims to oversee the development of these conditions, ensuring a balanced approach to leveraging opportunities and mitigating risks. Limitations of AI In the realm of artificial intelligence, responsible usage is paramount, yet it comes with inherent challenges. These systems, reflective of the biases and assumptions present in their training data and design, often mirror societal prejudices, influencing their decision-making processes. This phenomenon underscores the critical need for vigilance in the AI development phase to avoid encoding negative assumptions. Furthermore, AI lacks the intrinsic ability to evaluate its effectiveness, operating within a narrow set of pre-defined goals without the capacity for human-like judgment. This limitation highlights the fallacy of perceiving AI as capable of independent thought or impartiality, necessitating continuous human oversight to ensure alignment with broader societal values. Additionally, the foundation of AI on mathematical principles constrains its understanding to quantifiable metrics, struggling to comprehend the qualitative nuances that define human objectives. In summary, these limitations underscore the need for careful consideration in the design, deployment, and oversight of AI systems to ensure they serve human purposes without perpetuating existing biases or overlooking the complexity of human values. (Leal, et al., 2022) AI governance Artificial Intelligence (AI) not only reflects inherent biases but also significantly impacts societies based on the context it's deployed in. Local governments must understand how values are integrated into AI, steering its development towards inclusivity and sustainability. This involves a deep dive into AI governance, which is not limited to digital oversight but includes ethical guidelines, legal regulations, societal norms, and practices. Governance here is a broad concept, designed to capture various interpretations and focus on decision-making and social interactions. (Leal, et al., 2022) Despite national AI guidelines, cities face challenges within a multi-level governance system, highlighting the need for a nuanced approach that allows them to leverage their unique positions. (Schmitt, 2022) (Taeihagh, 2021) AI also poses accountability challenges, necessitating robust frameworks to balance risk and opportunity throughout the AI lifecycle. A significant hurdle is the limited capacity of cities to navigate the AI governance landscape, amid a high demand for IT and AI specialists. This challenge can be addressed through cross-sector partnerships and a focus on developing local talent, ensuring that cities' core values guide AI initiatives towards a more inclusive and sustainable future. For that reason, the European Commission has developed the first-ever legal framework on AI, the AI Act. This initial proposal, crafted by the European Commission, was designed to provide clarity and guidance to AI developers and users while minimizing burdens, particularly for small and medium-sized enterprises (SMEs). Key elements of the proposal included a governance structure at both European and national levels to address AI risks and a risk-based classification of AI systems into four categories. The strictest regulations were reserved for AI practices deemed to pose unacceptable risks, such as social scoring by governments. High-risk applications, notably those in critical infrastructure or law enforcement, were subject to rigorous pre-market requirements including comprehensive risk assessments, data quality control, traceability, and mandatory human oversight. Conversely, AI systems with limited risk were mandated to maintain transparency to enable informed public interactions, such as clear identification of AI-generated content and chatbot disclosures. The least risky applications, like video games or spam filters, were allowed more operational freedom, reflecting their benign usage across the EU. On March 6, this foundational proposal transitioned into enforceable law as the European Parliament approved the Artificial Intelligence Act. The legislation, which had been refined through negotiations with member states since December 2023, was robustly supported by MEPs, passing with 523 votes in favor, 46 against, and 49 abstentions. The ratified Act continues to protect fundamental rights, democracy, the rule of law, and environmental sustainability, shielding them from the risks associated with high-risk AI technologies. It aims to foster innovation and secure Europe's leadership in AI by setting tailored obligations based on the risk and impact level of different AI systems. The Act specifically prohibits AI applications that could infringe on citizens' rights, such as biometric categorization based on sensitive characteristics, indiscriminate scraping of facial images for recognition databases, emotion recognition in workplaces and schools, social scoring, profiling-based predictive policing, and any AI that manipulates human behaviour or exploits vulnerabilities. Law enforcement use of biometric identification systems, although generally prohibited, is permitted under strictly defined and narrowly applicable circumstances, such as locating a missing person or preventing terrorist attacks, subject to stringent temporal, geographical, and approval constraints. High-risk AI systems now face stringent obligations to evaluate and mitigate risks, maintain transparency, ensure human oversight, and log usage. General AI systems are required to adhere to transparency standards, including compliance with EU copyright laws and the publication of detailed summaries of their training content. Noteworthy innovations include the introduction of labelling requirements for manipulated content, known as "deepfakes", and the establishment of national-level regulatory sandboxes to foster the development of innovative AI technologies. As the Act moves towards formal adoption and implementation, it embodies the principles and recommendations from the Conference on the Future of Europe, aiming to create a competitive, safe, trustworthy society with transparent and responsible AI usage, while improving digital accessibility for all, including those with disabilities. This legislative milestone marks the beginning of a new era in AI governance, aligning closely with European values and strategic ambitions. (European Commission, 2024). The AI Life Cycle The development of an artificial intelligence (AI) system is segmented into five interconnected stages that outline its interaction with the external environment. This model aids in comprehending AI's architecture and its progression, promoting strategic analysis. The stages form a cycle, each linked to the others, influencing aspects ranging from algorithm creation to final deployment. Beginning Phase The journey starts with the beginning phase, focusing on identifying the problem. This initial step lays the foundation for the rest, as everything that follows is tied back to the original problem identified. It includes critical assessments and identification of potential risks in the AI system's intended use. Design Stage Next, the design stage emphasizes planning the algorithm's structure before any coding begins. It builds on the groundwork laid in the beginning phase, taking into account team dynamics and the potential impact of introducing AI, such as changes in power dynamics and economic impacts. Technical Creation Stage The technical creation stage shifts towards the practical aspect of AI development, concentrating on the specifics of the algorithm and the technical decisions involved. Implementation Phase After development, the implementation phase transitions the algorithm from a controlled setting to its real-world application. This phase is critical as it exposes the algorithm to the complexities of its operational environment. Maintenance Phase Lastly, the maintenance phase focuses on the period after the AI system has been deployed, up until it is decommissioned. This phase is crucial for ensuring the algorithm's continued effectiveness and relevance through regular updates and maintenance. In urban contexts, AI systems typically function within a complex ecosystem, necessitating proactive risk identification and management. Integrating AI development stages with local project management methods offers strategic opportunities for intervention and adaptation, ensuring the AI systems align with urban dynamics and challenges. (Leal, et al., 2022) Stages of implementation of AI solutions in smart cities Understand the Difference Between AI and ML for Smart Cities: Begin by educating stakeholders on the nuances of AI and ML. Understanding these technologies' capabilities is crucial for identifying how they can be applied to enhance urban living, such as through traffic flow optimization or predictive infrastructure maintenance. Define Urban Challenges and Opportunities: Clearly identify the specific problems or opportunities within the city that AI can address. Ask critical questions about desired outcomes, existing obstacles, the role of AI in overcoming these challenges, and the data available to support these efforts. Prioritize Value-Driven AI Initiatives: Select AI projects based on their potential to deliver tangible benefits to the city and its residents, focusing on near-term goals. This might include improving public transportation efficiency or enhancing energy sustainability in public buildings. Evaluate Capabilities and Approach for AI Development: Assess whether to develop AI solutions in-house, purchase existing technologies, collaborate with external partners, or outsource development, based on the city's internal capabilities and strategic objectives. Consult with Smart City AI Specialists: Engage with domain experts who have experience in applying AI within urban environments. This could include academic researchers, private sector innovators, and technologists from other smart cities. Prepare and Secure Urban Data: Organize and clean the city's data to ensure it is ready for AI applications. Invest in robust security measures to safeguard this data, respecting privacy and ethical considerations. Start Small and Scale Thoughtfully: Launch pilot projects targeting specific urban challenges to demonstrate the potential of AI. Use the insights gained from these initial efforts to guide the strategic expansion of AI applications across different city domains. (Majewski, 2023)