Why AI ethical laws, governance and regulations are important and should remain relevant

Introduction

Our lives evolve, and so does everything that applies directly to human lives. Artificial Intelligence (AI) has become part of our lives so, AI needs to have protective guardrails (Ethical Laws, Governance, and Regulations) to ensure our safe usage of the AI systems.

💡

Throughout this article, I use both the doctor-patient analogy and the patient-digital health AI analogy.

Think of a time when you felt sick, paid your doctor a visit, and during the consultation, you were asked a series of questions to assess your medical situation. Your doctor, who is a highly trained medical professional working with you as your medical assistant to provide the relevant medical support, needed to get more information about the context of your illness and general information about your health to determine the suitable next steps to provide adequate and safe medical support for you. AI systems work similar to your doctor. Let us explore some fundamentals in AI to understand better.

What is AI?

AI is an acronym that means Artificial Intelligence. Just like your doctor, AI is a highly trained system that uses a vast amount of data, allowing it to act or mimic human-like intelligence to perform tasks based on the context of the data it receives. AI uses learned patterns and its vast amount of knowledge(data sources), like your doctor, to generate responses tailored to your specific context (inputs).

In simple terms, and according to Microsoft Azure, AI is defined as “the ability of a computer system to mimic human-like cognitive functions, such as learning and problem-solving” .

According to IBM, AI is “a branch of computer science dealing with the simulation of intelligent behavior in computers”.

Types of AI

AI has been around for over 50 years, with evolving abilities from the Checkers program in 1951 to recent applications like Large Language Models, example; GPT-4 and LLaMA 3, and others. Scientists have continuously contributed to improving the quality and ability of machines to mimic human intelligence. A brief history of AI dates to the early to mid-90s, and researchers have defined levels of AI that speak to its capabilities, categorized in three types. The three categories are:

1. Weak AI (or Narrow AI or Artificial Narrow Intelligence): This type of AI is designed to perform a specific task or a set of related tasks. It operates under a narrow range of conditions and lacks general cognitive abilities. Example, Siri, Netflix movie recommendations, etc.

2. General AI (or Artificial General Intelligence, AGI): This is a form of AI that can understand, learn, and apply intelligence across a broad range of tasks, like human cognitive abilities. Currently, AGI remains theoretical and has not yet been achieved.

3. Strong AI (or Artificial Superintelligence): This hypothetical AI would surpass human intelligence across all aspects, including creativity, problem-solving, and social skills. It remains a topic of speculation and future potential.

Foundations of AI

For machines to behave in certain ways they are preprogrammed to behave like they do, and this applies to AI. So, what are the core building blocks of AI that makes it function like it does?

AI systems require to be trained on vast amounts of data to enable them to learn and evolve in improving their responses and outcomes. This means, data is a powerhouse for the application of AI systems.

1. Data: For AI systems to behave in a predefined way, mimicking any of the five human aspects, they need adequate high-quality data for optimal development. Data is collected properly and securely, cleaned, and stored in a suitable data storage solution for preprocessing. The pre-processed data is fed to the machine learning model for training.

2. Preprocessing: Data must be pre-processed into formats that AI systems can understand, by vectorization and tokenization.

3. Machine Learning: Machine learning involves training algorithms to understand and learn from data inputs to make decisions or predict outcomes. There are three main types:

a. Supervised Learning: Involves the use of labelled data to train algorithms to predict results or classify items. For example, labelling drugs in an online pharmaceutical store.

b. Unsupervised Learning: Involves training models with unlabeled data to find hidden patterns or intrinsic structures. For instance, recognizing behavioral patterns of patients and grouping them according to recognized patterns in their behavior.

c. Reinforcement Learning: This involves an AI agent that learns by interacting with its environment and receiving feedback, which it uses to improve its outcomes and decision-making.

4. Algorithms: Algorithms are sets of well-defined instructions a computer uses to solve a defined problem. For example, your doctor followed a set of steps to diagnose your health condition and decide on the optimal medical support you need. Algorithms implement this step-by-step decision-making process.

5. Deep Learning: Is a subset of machine learning that enables Machine learning models to simulate the complex decision-making process of the human brain by using multilayers of neural networks (neural networks are series of algorithms taking inputs and transferring their outputs across layers within the algorithm). This enables the network to learn, model complex data patterns and make complex decisions.

6. Natural Language Processing (NLP): Is a subset of AI focused on providing a way for computer systems to understand, interpret and generate human language by combining machine learning, deep learning, and computational linguistics to process and understand human language.

Human intelligence AI aims to mimic.

AI is designed around the ability of machines to simulate human knowledge majorly around five areas;

1. Learning: As humans we acquire knowledge from our interactions. AI is designed to use machine learning, deep learning, and reinforcement learning to learn from interactions, much like humans learn from interactions.

2. Problem Solving: Like how humans apply known procedures and knowledge to decide on the best approach for solving a task, AI applies algorithms, heuristics, and optimization techniques to make predictions and solve complex problems.

3. Language Understanding: AI uses Natural Language Processing(NLP) to understand human language, predict patterns, and perform sentiment analysis, allowing it to recognize and respond to emotional cues.

4. Reasoning: AI uses predefined rules or algorithms and knowledge bases to make decisions and draw inferences, akin to how humans use their experiences, mental or mind maps, in brainstorming for decision making or prediction.

5. Perception: Each day, the human brain processes over 10gigabytes of data from our sensory organs allowing the brain to make better decisions. For an AI system to replicate this ability to perceive its environment, it uses computer vision and natural language processing to replicate sensory perceptions, processing and understanding visual and textual information.

Artificial intelligence — Machine Learning, Robotics, Algorithms | Britannica

Now we reviewed some of the fundamentals of AI systems, we would now explore the need for ethical laws and regulations.

Why we need ethical laws and regulations for AI systems

Like your doctor, who is governed by strict professional ethics and laws, to ensure safe medical practices, AI systems need to be guided by laws and regulations to ensure safe and secure usage. These laws are essential to protect privacy, promote fairness, ensure transparency, prevent bias, maintain accountability, and build trust. Without such systems in place, AI could be misused, leading to societal harm, including monetary, physical, economic damage, and even potential wars.

There are various AI laws implemented for the safeguarding the development and adoption of AI systems. For this article, I will be focusing on the UNESCO Ethical AI framework and Microsoft’s AI Principles.

Ethical AI Frameworks

The UNESCO Ethical AI Framework is the first global AI framework adopted by all 193 member countries in November 2021. It aims to implement strategies to uphold human dignity, protect human rights, and guide the ethical development and deployment of AI technologies. It addresses critical issues such as transparency, accountability, privacy, and data protection. Below are key points for each of the ten principles stated in the document:

1. Proportionality and Do No Harm: “None of the processes related to the AI system life cycle shall exceed what is necessary to achieve legitimate aims or objectives and should be appropriate to the context.”

2. Safety and Security: “Unwanted harms (safety risks), as well as vulnerabilities to attack (security risks) should be avoided and addressed, prevented, and eliminated throughout the life cycle of AI systems.”

3. Fairness and Non-Discrimination: “AI actors should promote social justice and safeguard fairness and non-discrimination of any kind in compliance with international law.”

4. Sustainability: “The development of sustainable societies relies on the achievement of a complex set of objectives on a continuum of human, social, cultural, economic, and environmental dimensions.”

5. Right to Privacy, and Data Protection: “Privacy, a right essential to the protection of human dignity, human autonomy, and human agency, must be respected, protected, and promoted throughout the life cycle of AI systems.”

6. Human Oversight and Determination: “Member States should ensure that it is always possible to attribute ethical and legal responsibility for any stage of the life cycle of AI systems.”

7. Transparency and Explainability: “The transparency and explainability of AI systems are often essential preconditions to ensure the respect, protection, and promotion of human rights, fundamental freedoms, and ethical principles.”

8. Responsibility and Accountability: “AI actors and Member States should respect, protect, and promote human rights and fundamental freedoms and should also promote the protection of the environment and ecosystems.”

9. Awareness and Literacy: “Public awareness and understanding of AI technologies and the value of data should be promoted through open and accessible education, civic engagement, digital skills, and AI ethics training.”

10. Multi-Stakeholder and Adaptive Governance and Collaboration: “Participation of different stakeholders throughout the AI system life cycle is necessary for inclusive approaches to AI governance, enabling the benefits to be shared by all, and to contribute to sustainable development.”

Microsoft AI Principles

The Microsoft AI Principles emphasizes six core areas when developing AI systems. To remember these principles, I’ve coined the acronym PITFAR and attached it to a memorable food sentence(as a gourmand, for things like this I think food first - lol) — PleaSe Include Tasty Fries And RibS. Let’s elaborate on each of them:

1. Privacy and Security: AI systems should be designed to protect user privacy and secure data.

2. Fairness: AI systems should treat all people fairly and avoid biases that could lead to discrimination.

3. Reliability and Safety: AI systems should perform reliably and safely, with robust protections against unintended consequences.

4. Inclusiveness: AI models perform well for all users, and no one is excluded from the opportunities provided by intelligent solutions.

5. Transparency: AI systems processes should be understandable, and their decisions explainable to every user.

6. Accountability: The people who design and deploy AI systems must be accountable for how their systems operate.

The above guardrails are plausible examples of well-defined rules that any developer of an AI system should follow. They can also serve as a means for enforcing AI governance across systems. Governments around the world are developing more tailored ethical laws, and as these laws are implemented, the required legal penalties should be clearly stated for defaulters. These principles, laws, and regulations will ensure the safe development of AI systems and facilitate the enforcement of AI governance practices.

Typical ethical issues that may arise in AI.

Imagine you are using a digital health AI system to assist you with your diagnosis and treatment. Even though the technology is advanced, there are some tricky ethical issues that could come up:

1. Bias and Discrimination: Just like a doctor could unintentionally make biased decisions based on incomplete information or subconscious bias, AI systems can also reflect and even amplify biases present in their training data. For example, if an AI tool is used for medical diagnoses, it might not be as accurate for all demographics if its training data wasn’t diverse enough. This could lead to unfair treatment of certain groups or potential medical hazards.

2. Privacy Concerns: When you use your digital health AI system, it may involve sharing a lot of personal health data. If this data isn’t protected properly, there’s a risk of breaches or unauthorized access. Just as you trust your doctor to keep your information confidential, we need to ensure that AI systems safeguard your privacy.

3. Accountability and Responsibility: If the AI system makes a mistake in your treatment plan, it can be challenging to figure out who is responsible. Is it the AI, the developers who created it, or you the user? Clear guidelines are needed to address mistakes and ensure accountability. Developers of AI systems must provide for all users including stakeholders and co-developers, clear details on how their AI systems work and state in clear terms the responsibility for all parties involved.

4. Autonomous Decision-Making: Imagine your doctor started using AI to make treatment decisions without any human oversight to the outcomes of its decision. This could be a concern if the AI is making critical health decisions on its own. It’s important to ensure that AI systems are used responsibly and with appropriate human oversight and the outputs of Ai systems are reviewed for validate the accuracy.

5. Lack of Transparency: If an AI system is involved in your diagnosis but doesn’t explain its reasoning, it can be hard for you to understand how decisions are made. Transparency is key to ensuring trust and understanding in AI-assisted healthcare.

Why we need the laws and why we need to stay current

So, why is it important to have laws and keep them current stay up to date with them?

1. Adapting to Rapid Technological Changes: Like your doctor needs to stay current with the latest medical advancements, AI laws and regulations need to be up-to-date to the rapid developments in AI. This helps manage new risks and opportunities effectively, ensuring that AI tools are up-to-date and relevant.

2. Protecting Human Rights: In the same way medical ethics ensure your rights and dignity are respected during treatment, AI laws must protect human rights such as privacy and freedom from discrimination. Ensuring these rights are safeguarded in AI applications is essential for the safe usage of AI systems.

3. Maintaining Public Trust: Trust is crucial in the doctor-patient relationship, and the same goes for AI. Transparent and current regulations help build trust that AI systems are being developed responsibly and ethically. This gives the public confidence in the outputs of AI systems.

4. Encouraging Ethical Innovation: Just as your doctor follows ethical guidelines in patient care, more relevant and clear laws and guidelines must be implemented for ethical AI development. This encourages innovations that are aligned with societal values and ethical standards.

5. Preventing Misuse and Harm: Proper regulations are like safety nets that prevent misuse of AI and mitigate potential harms. They help ensure that AI systems, like those used in healthcare, are secure and used responsibly, avoiding negative impacts on patients and society. Therefore, AI governance cannot be overlooked as it is vital for enforcing these AI guardrails.

Conclusion

Understanding AI’s capabilities, types, and the importance of ethical regulations is crucial as we develop and integrate these technologies into society. By adhering to ethical frameworks and principles, we can ensure AI systems are developed and used responsibly, safeguarding human dignity and promoting fairness, transparency, and accountability.

💡

If you enjoyed reading this or need to discuss this article, reach out to me on Twitter. It does not end here; see below for other details. Thank you for reading this article, and I hope you enjoyed it.

Some other AI laws and frameworks

Glossary

GPT-4 generated content

Heuristics: Simple rules or shortcuts used to make quick decisions or solve problems without needing a detailed analysis.
Computational Linguistics: The study of how computers can understand and work with human language using algorithms and models.
Models: Mathematical tools or algorithms that learn from data to make predictions or perform tasks, like recognizing patterns.
Data Storage Solution: Systems used to keep and manage data, such as hard drives, databases, and cloud storage services.