Hello, World: Artificial Intelligence and its use in the public sector
AI for government administrations by OECD Observatory of Public Sector Innovation publication from 2019
A few years back, I helped a group of experts from the OECD Observatory of Public Sector Innovation write a beginner’s guide on Artificial Intelligence [Hello, World: Artificial Intelligence and its use in the public sector]. At that time, not many people knew about it, and some public offices didn’t take it seriously. But now, with the quick growth of AI like ChatGPT and other new tools coming out all the time, the public sector is paying more attention. Here’s a short summary of the 2019 report, inviting everyone working in public services to learn more about AI (it’s a summary of a fragment, the whole report is extensive). There’s a good chance a new beginner’s guide will come from OECD soon. Keep an eye out for it!
Hello, World: AI in the public sector
Artificial Intelligence (AI) holds great potential for the public sector, with governments having the ability to set priorities, investments, and regulations. Understanding AI is crucial for policymakers and civil servants to determine if it can help achieve their goals. AI currently refers to machine-based systems that make predictions, recommendations, or decisions. Governments need access to sufficient, unbiased data to take full advantage of AI.
AI can be classified as “narrow AI,” which is used for specific tasks such as natural language processing and computer vision. Machine learning techniques hold significant potential for various tasks but have their own strengths and limitations. Governments are looking to catch up with the private sector in AI adoption, with many countries launching or planning national AI strategies.
AI can promote innovation in the government sector but is not a solution for every problem. Governments should carefully consider whether AI is the best solution for a specific issue and take into account various factors when exploring AI, including supporting experimentation, ensuring multidisciplinary perspectives, developing ethical frameworks, securing data, and providing resources for AI implementation. This primer discusses different approaches to AI innovation that can be adapted for use in various countries and contexts.
Defining Artificial Intelligence
Defining Artificial Intelligence is a complex task due to the subjective nature of the term “intelligence” and the constantly evolving perception of AI capabilities. The artificial aspect of AI is clear; it refers to non-natural, man-made machines, computers, or systems. However, many definitions and perspectives exist for AI, as understanding what constitutes intelligence varies.
John McCarthy, the father of AI, defined it as “the science and engineering of making intelligent machines.” Alan Turing’s influential approach to defining AI involved comparing machines to humans in displaying intelligence. The OECD definition of AI systems is accepted by 42 national governments and focuses on the creation of algorithms that can learn and reason.
Different organizations and sectors offer their own definitions of AI, such as the European Commission’s High-Level Expert Group, the Luxembourg regulation authority, the UK government, and the IEEE Standards Association. AI can be understood as the field of knowledge associated with the design of these machines or as specific capabilities within a machine, such as writing music or solving math equations.
Perceptions of intelligence change over time, with some tasks previously considered AI now being seen as automation. The AI Effect describes this phenomenon of AI capabilities becoming normalized. The primer aims to provide an overview of AI and its potential applications and considerations for public sector innovation and transformation by exploring the current intelligence of machines.
AI can generally be categorized into two broad perspectives: General AI and Narrow AI. General AI, also known as strong AI or Artificial General Intelligence (AGI), refers to machines matching or surpassing human intelligence across domains and capabilities. Although some experts believe General AI may be possible in the future, it currently remains in the realm of science fiction.
Narrow AI, also known as weak AI, applied AI, or Artificial Narrow Intelligence (ANI), reflects the current state of AI, where algorithms and machines can outperform humans in specific tasks but not across a wide range of tasks. Examples of Narrow AI subfields include computer vision, natural language processing (NLP), speech recognition, knowledge-based systems, and automated planning. These subfields can evolve and overlap to create new applications.
Artificial Intelligence Augmentation or Intelligence Augmentation emphasizes the interaction between humans and machines, allowing them to collaborate and solve problems more effectively than either could do alone. This approach is important for human-computer interaction design and can be applied to the public sector to augment the abilities of civil servants, enabling them to process large amounts of data and provide better services to citizens.
In conclusion, while General AI and Narrow AI reflect different outlooks on AI’s potential, the current state of AI, centered on Narrow AI, presents both opportunities and challenges that need to be addressed. Governments should explore immediate AI opportunities while preparing for potential long-term technological shifts.
Intelligence augmentation: Examples of human-machine collaboration
Intelligence augmentation emphasizes human-machine collaboration, particularly in creative tasks. One approach to this collaboration involves evolutionary and genetic algorithms, where a problem is defined, and potential solutions are generated. Human interaction is incorporated by selecting from these potential solutions, with the AI learning from the choices made.
An example of this approach is Sung-Bae Cho’s AI tool that generates dress designs. Users select which designs to keep, and the AI learns from these choices, evolving to generate new designs. Similar systems have been developed in industrial engineering, medicine, and video games, providing creative solutions for real-world problems.
Generative interfaces enable human users to interact with and understand all potential generated solutions, fostering human-machine collaboration. This collaboration doesn’t replace human creativity but augments it, allowing humans to explore thousands of unique solutions quickly and efficiently.
Renewed enthusiasm for AI
AI has experienced periods of enthusiasm and disinterest, metaphorically referred to as “winters” and “springs.” Currently, AI is in an “AI spring” due to several factors:
- Maturity of the field: Accumulated knowledge, refined algorithms, models, programming languages, and frameworks have led to the development of new AI applications. Deep Learning, for example, emerged only in the last decade.
- Better technology: Modern computers are more affordable, powerful, and compact, allowing for faster processing and larger, more complex programs. Data storage costs have also decreased dramatically.
- Democratization of computers and programming: Technology is more accessible to a wider range of users, and collaborative platforms and tools enable programming for individuals from diverse backgrounds. Free online courses and tutorials further contribute to this democratization.
- Data availability and machine learning: The increasing volume of data generated by daily digital interactions, often referred to as “Big Data,” has driven the development of AI applications. This abundance of data, coupled with advanced processing technologies, has fueled the progress of Machine Learning and Deep Learning.
These factors contribute to the ongoing and growing optimism surrounding AI in both the private and public sectors.
What is next for AI? [from 2019]
The future of AI remains uncertain, with rapid developments impacting various aspects of society. AI systems have demonstrated significant results in areas such as medicine and self-driving cars, but a cautious approach is necessary due to the potential for unrealistic expectations and hype.
In the public sector, the situation is less clear, with less thought and research devoted to how emerging technologies will impact governments. Gartner’s Hype Cycle for Digital Government Technology predicts that Machine Learning may take only 2–5 years to reach a fully productive state. However, the history of AI suggests that advancements will continue, regardless of periods of disillusionment.
Governments must understand AI and its potential effects on the public sector to make informed decisions about its use. Some governments are already benefiting from AI, but there is little room for a “wait and see” approach, as failing to keep up with technology may undermine their ability to tackle increasingly complex issues. It is crucial for public organizations to engage with AI and its underlying technologies, considering their implications on institutions and citizen interactions. This will help public servants become familiar with AI’s capabilities and prepare their organizations to leverage the technology responsibly.
