2024 Public Sector Perspectives

However, the technology that underlies such chatbots is relatively unsophisticated compared to what is possible with generative AI. Early chatbots applications cannot respond to complex queries or questions: when they encounter a problem, the chat is usually transferred to a human operative. Today’s generative AI-chatbots can leverage sophisticated natural language processing (NLP) techniques via LLMs to produce human- like text, making interactions more context-aware (increasingly, these chatbots are beginning to sound like humans too). Traditional AI techniques ingest data to learn how humans behave and analyze this information to replicate patterns, effectively enabling it to predict what happens next. Generative AI goes a step further by analyzing existing data from a variety of inputs to generate new content. Crucially, generative AI-chatbots can integrate with various data sources via application programming interfaces (APIs) or plugins, enabling them to retrieve real- time information and performmore complex tasks. As well as providing access to information and services, generative AI might be used by governments to solicit, analyze and action citizens’ opinions on various topics – ranging from identifying and verifying locations of potholes in a neighborhood to what government should prioritize in order to reach net zero emissions by 2050. Crowdsourcing gives governments a new way to understand the opinions of its citizens and more rapidly respond to their concerns with policies and actions. Generative AI can be used to more rapidly collate and act upon data gathered from citizens. In Belgium, civic tech company CitizenLab uses NLP-powered AI to process and categorize thousands of citizen contributions, highlighting key trends for more efficient decision-making. Civil servants can access this data via real-time dashboards, enabling them to easily identify citizens’ priorities and to make decisions accordingly. For example, in 2019 when large numbers of young people were protesting against climate change, CitizenLab set up a participation platform that generated 1,700 ideas, 2,600 comments and 32,000 votes on the topic. These findings were used to inform a report for elected officials that included 16 policy recommendations. 2 Operational enablement Many government bodies have multiple manual touch points in their operations. AI, and specifically generative AI, can be used by public sector entities to enhance processes, connect data points together, and automate time-consuming tasks such as reviewing data and submitted documents, such as tax returns and regulatory reports. Generative AI can be used to analyze vast unstructured datasets (such as emails, phone calls, regulatory reports, news articles, social media posts, images and videos) in more detail than was previously possible in order to identify trends, patterns, and potential problems, and to generate insights that can be used for better decision-making and policy-making. 2 https://oecd-opsi.org/wp-content/uploads/2019/11/AI-Report-Online.pdf Potential barriers to AI adoption While there are many benefits from adopting AI, government and public sector entities face some challenges in introducing new technologies: • AI have the potential to lower costs but may require significant upfront investment. • The public sector is in competition for AI talent with the private sector, whichmay be able to offer more generous financial rewards. • The scale of government can sometimes result in a culture of risk aversion and hesitancy to innovate in advance of the private sector. • While government has huge volumes of valuable data, it may not be easily accessible or in usable formats: government IT infrastructure may lag the private sector. For example, it might be used in healthcare to predict disease outbreaks by aggregating and analyzing data from electronic health records, public health databases, environmental data (such as weather and climate information), social media, and even internet search trends. Australia’s Epiwatch uses NLP to search for phrases that could be associated with the emergence of a new illness, for instance. 3 Generative AI can potentially take this analysis further by ‘understanding’ the meaning behind more complex phrases and adjusting for differences in local dialects. Similarly, if done correctly, law enforcement can use AI to predict crimes by analyzing historical data, including the time and location of past crimes, or by continuously monitoring real-time data such as CCTV feeds and social media posts. On a larger scale, AI is being used to fight organized crime in remote areas of Africa. EarthRanger, part of the Allen Institute for Artificial Intelligence, uses AI and predictive analytics to collate historical and real-time data – including from ranger patrols, spatial data and observed threats – within protected areas. The technology has helped park rangers to crack down on poaching in the Grumeti Game Reserve, Tanzania. 4 Again, generative AI has the potential to supercharge this analysis by analyzing vast amount of unstructured data that can further provide support for a particular strategy or immediate action. In transportation, AI can improve traffic management by optimizing traffic flow, reducing congestion and improving road safety. Again, this process can operate in real time by leveraging smart traffic lights and adaptive speed limits. In the UK, AI is now being trialed for air traffic control using simulations of real-life air traffic, although its introduction is some years away. 5 In a similar way, California’s firefighters now use AI to analyze data from more than 1,000 cameras across the state to spot wildfires and mobilize first responders. 6 3 https://harvardpublichealth.org/disease/ai-services-like-epiwatch-are-already-tracking-the-next-pandemic/ 4 https://enactafrica.org/enact-observer/ai-and-organised-crime-in-africa 5 https://www.ft .com/content/783a9d91-cce3-4177-bfe0-5438aa3b892a 6 https://www.reuters.com/world/us/california-turns-ai-help-spot-wildfires-2023-08-11/ 7 https://www.weforum.org/agenda/2023/08/artificial-intelligence-techonology-news-august/ Needless to say, the new generation of AI techniques, including generative AI, promises to further provide assistance when reviewing complex scenarios and situations that might arise on-the-fly. Risk and controls Government and public sector bodies have a huge impact on their country’s economy and society and, as such, require effective risk management and controls. AI can strengthen their capabilities and – given the speed that data can be analyzed – could be used to design risk policies and controls that respond dynamically as circumstances change. One obvious government use of AI is to enhance cybersecurity by detecting and responding to cyber threats in real-time, protecting sensitive government data and critical infrastructure. In August, the Biden Administration launched a competition to challenge private sector companies to use AI to identify and resolve software vulnerabilities in areas from critical infrastructure to the code that runs the internet. 7 Generative AI can be used to analyze vast unstructured datasets (such as emails, phone calls, regulatory reports, news articles, social media posts, images and videos) in more detail than was previously possible in order to identify trends, patterns, and potential problems, and to generate insights that can be used for better decision-making and policy-making. Citi Perspectives for the Public Sector 33 32 AI in the Public Sector: Facilitating Engagement, Improving Enablement and Enhancing Risk Management