7 challenges of artificial intelligence and possible uses in public administration

How can public administration use artificial intelligence and new technologies? The associated risks and benefits.

Defining artificial intelligence

Artificial intelligence (AI) is a discipline concerned with studying and creating computer systems capable of emulating behaviours and processes typically associated with human intelligence. These systems use complex algorithms, machine learning, artificial neural networks and other techniques to perform tasks such as image recognition, natural language understanding, decision-making and problem-solving.

Note: this definition was generated by a generative AI system.

The term artificial intelligence can itself cause misunderstandings. At present, so-called AI is not intelligent in the traditional sense. It is a collection of data and algorithms that can emulate intelligence fairly well.

Any sufficiently advanced technology is indistinguishable from magic.

The current situation

We are in a golden age for AI, especially generative AI. There has been much discussion of ChatGPT, DALL-E and others. These systems generate output from a question written in ordinary language, called a “prompt”. Output can be almost anything: text, images, music, voice, video. For every need, there is an application and an associated model.

These models are often open source, allowing anyone with some technical familiarity to adapt them to generate more specific content.

Uses

Generative AI applications are already used in a wide range of fields:

• Creating content, including newspaper articles.
• Refining product design, for example at Ferrari.
• Creating entire marketing campaigns.
• Market analysis.
• Customer support.
• Translation.
• Transcription.
• Accessibility, such as speech synthesis.
• Help with software development, such as GitHub Copilot.

There are already too many established uses for an exhaustive list.

Challenges

Some of the most obvious challenges of AI, especially generative AI, include:

1. Output quality: generative AI can produce results of varying quality. Accuracy and consistency may be problematic, especially in professional contexts where precision is essential.
2. Ethics and bias: generative AI can reflect cultural or social biases, producing discriminatory or one-sided content. Managing ethics and bias is a major concern.
3. Surveillance and privacy: generative AI raises privacy concerns, especially when generating content involving personal or sensitive information.
4. Copyright and intellectual property: AI-generated content can raise copyright and ownership questions, especially when based on protected data. Several companies have already said they are prepared to defend customers facing intellectual property claims, starting with Microsoft, followed by Adobe and Google.
5. Cost and accessibility: implementing and using generative AI may require significant investment in hardware, software and staff training, creating a barrier for some organisations.
6. Accuracy and responsibility: decisions based on AI output can carry risks and liabilities. Understanding its limits and ensuring appropriate human oversight is essential.
7. Regulation and compliance: AI is evolving, so regulation may be unclear or incomplete. Users must monitor relevant laws and rules and ensure compliance.

Public administration

Public administration should seize the opportunities these systems offer, while always considering the challenges they bring. They can improve public services, interactions between citizens and local authorities, and the efficiency of administrative and management processes.

• Translate websites and documents for linguistic minorities and foreign residents.
• Create systems through which citizens can request information or documents.
• Analyse large healthcare datasets quickly, as happened with CT scans in some hospitals during the COVID pandemic.
• Help tourists find relevant information, accommodation, events and cultural activities.
• Improve safety through active video surveillance that identifies dangerous situations and alerts a control centre.
• Improve road maintenance by detecting problems through cameras or other sensors.
• Automate processes to free resources for better public services.
• Help study traffic-management solutions.
• Improve traffic through smart, centrally managed lights and signs.
• Optimise separate waste collection.
• Improve road safety with sensors detecting dangerous situations such as wrong-way driving, incorrect entrances, prohibited overtaking, access to restricted areas, calls for help or unusual vehicle stops.
• Save energy and reduce environmental impact in street lighting management.

This list is also far from exhaustive: the applications of AI and new technology are almost endless. Introducing some improvements would free resources that could be trained to study further solutions, creating a virtuous circle that makes public administration more efficient.