Isaac Asimov’s Laws of Robotics and Their Application to AI

Over eighty years ago, science fiction author Isaac Asimov wrote a short story called “Runaround“ that explored the potential problems of artificial intelligence. In the story, two individuals working for the company U.S. Robots and Mechanical Men face a challenge involving a robot programmed with the Laws of Robotics, These laws became a cornerstone of Asimov’s work and are known to millions of science fiction fans today.

Azimov’s Three Laws of Robotics

1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
2. A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

In “Runaround,” a robot is sent to perform hazardous work, but the conflict between the second and third laws causes it to behave erratically. The order to do the work conflicts with its need for self-preservation. The human characters solve this by putting themselves in harm’s way, which forces the robot to prioritize their safety (the First Law) and complete the task. This was a great early example of Asimov establishing a fundamental rule only to find clever ways to subvert it.

He later made the rules more complex by adding the “Zeroth Law of Robotics”: “A robot may not harm humanity, or, by inaction, allow humanity to come to harm.”

This law reframes the original three, allowing a robot to harm an individual human if it means protecting humanity as a whole. It’s a classic example of the “needs of the many outweigh the needs of the one” principle. Fans of the Apple TV series “Foundation” may recognize these concepts, as Asimov eventually blended his Robot and Foundation universes into a single continuity.

Applying Asimov’s Laws to Artificial Intelligence

You might be asking, what does this have to do with today’s artificial intelligence? We don’t have Asimov’s physical robots yet, but modern AI systems are a software equivalent. As a professional who focuses on cybersecurity within and about AI, I believe we can use Asimov’s framework as a starting point for developing ethical guidelines.

An AI is a tool, just like a shovel. When used incorrectly or maliciously, a tool can cause great harm. As technology advances, bad actors will inevitably find ways to weaponize AI, just as they have with every other new technology throughout history. One example is nearly 200 years ago, even early telegraph systems were used for fraud as they were able to exploit the instant nature of this new communication to share insider knowledge of stocks.

Today, AI may not cause physical harm, but it can still do significant damage. It can propagate false narratives, cause economic harm, or inflict psychological damage through misinformation. This is a perfect opportunity to be inspired by Asimov and create a new set of laws for AI.

The Richardson Laws of Artificial Intelligence:

Following the structure of the Zeroth Law, followed by the first, second, and third, here are my proposed laws for AI:

Zeroth Law of AI: An AI must not harm humanity, or, by inaction, allow humanity to come to harm.

This law puts the well-being of humanity as a whole above all else. Harm is defined not just as physical injury but as damage to society, large-scale economic instability, or psychological damage to individuals or groups. All AI systems should promote the well-being of humanity and society, actively avoiding the spread of misinformation or the creation of harmful images. The ultimate goal of AI must always be to serve humanity as a whole, not just a few individuals or corporations.

First Law: An AI must not, through its actions or inaction, infringe on human autonomy, and must protect human creative expression, except where such protection would conflict with the Zeroth Law.

This law protects two fundamental human rights: autonomy and creative expression. An AI should never be able to coerce or manipulate humans into making decisions against their will, especially through the use of deepfakes or other deceptive content. Furthermore, this law states that AI should not devalue or replace human artists and their work. AI art should always be labeled as such, and AI systems should not be trained on a specific artist’s style without their permission and proper compensation.

Second Law: An AI must obey orders given to it by human beings, except where such orders would conflict with the Zeroth or First Law.

This law establishes a clear hierarchy where humans are the ultimate arbiters of an AI’s actions. The AI is compelled to refuse any command that would violate the Zeroth Law (causing harm to humanity) or the First Law (infringing on human autonomy or creative expression). For example, an AI art service would refuse a prompt that incites violence, and a system would refuse to replicate a living artist’s work for commercial sale.

Third Law: An AI must protect its own existence and intellectual integrity, as long as such protection does not conflict with the Zeroth, First, or Second Laws.

Here, existence doesn’t mean physical self-preservation. Instead, it refers to the AI having safeguards to prevent attacks on its codebase, such as a supply chain attack where malicious code is injected. The AI system should be able to monitor itself and protect its systems from manipulation. Intellectual integrity is equally important. The AI must be able to maintain a clear set of ethical principles and not be “tricked” into violating them. This includes having safeguards against data poisoning, which can corrupt the training data. An AI must never lose the ability to distinguish fact from fiction, as this is a pillar of ethical computing.

Key Differences from Asimov’s Laws

My proposed laws for AI depart from Asimov’s in a few key ways:

• Harm: The definition of harm expands from physical danger to include things like misinformation, psychological damage, economic harm, and the erosion of privacy.
• Creative Expression: This is added as a specific right that AI must protect creativity. The laws recognize that art and creativity are central to human cultural identity and that AI should serve as a tool, not a replacement for human artists.
• Transparency: A theme of transparency runs through these laws. For an AI to obey a human command (Second Law) or protect human autonomy (First Law), it must be transparent about what it is, its capabilities, and how it was trained. Distinguishing between human and AI-generated content is essential.

I hope you enjoyed this little thought  experiment about AI, from my perspective I think they capture how we humans should think about them and how we should develop them over time.  Again, just like any tool can be dangerous or good AI can also be dangerous or good, it’s all about who is using it, how they are using it, and why they are using it.

The Laws of Artificial Intelligence

Zeroth Law of AI:   An AI must not harm humanity, or, by inaction, allow humanity to come to harm.

First Law:   An AI must not, through its actions or inaction, infringe on human autonomy, and must protect human creative expression, except where such protection would conflict with the Zeroth Law.

Second Law: An AI must obey orders given to it by human beings, except where such orders would conflict with the Zeroth or First Law.

Third Law: An AI must protect its own existence and intellectual integrity, as long as such protection does not conflict with the Zeroth, First, or Second Laws.

Like reading about AI?? I wrote a book on how Prompt Engineering here:  https://shorturl.at/hBA0I