“An AI may not be used to harm humanity, or, by not being used, allow humanity to come to harm.”
As Artificial Intelligence systems (AI) like #ChatGPT enter into our lives and our work, we need some basic guidelines for how we implement them going forward. Here’s a place to start:
The Zeroth Law of AI Implementation:
An AI may not be used to harm humanity, or, by not being used, allow humanity to come to harm. Implement AI in such a way that you treat humanity, whether in your own person or in the person of any other, never merely as a means to an end, but always at the same time as an end, and treat AI always as a means to an end and never as an end in itself.
The sufficiently esoteric sci-fi and philosophy reader will note these are rewrites and recontextualizations of Isaac Asimov’s Zeroth Law of Robotics and the second formulation of Kant’s Categorical Imperative.
The Breakdown
My proposed Zeroth Law of AI Implementation aims to ground us in a shared purpose and vision for how we build our future with AI. It sets forth four basic principles:
- Do No Harm* with AI.
- Harm can be caused by having a tool and refusing to use it or otherwise limiting its use. For example harm can be caused by limiting access or capability based on factors including socio-economic status, geography, disability, etc.
- Humans are always ends in themselves, and must never be considered only means to an end (see “The Age of Surveillance Capitalism“).
- AIs are always means to (human) ends, and must never be ends in themselves.
* We need a clear definition what “harm humanity” means including 1) who gets to determine what constitutes harm, 2) who can be harmed, and 3) who adjudicates whether harm has been caused.
The Reason
The goal of technology is to build futures where humans have the capabilities to be and do what they have reason to value. AI technology presents an infinite possibility space for us to make that happen. AI technology also allows us potential to limit our own capabilities and cause real harm.
These statements seem self-evident, yet when technology is developed and implemented, we often forget its core purpose as we are blinded by the capabilities of the technology (it becomes an end in itself), by the power it affords us (those with access gain powers not afforded to others), and by the financial opportunities it affords us (using technology to turn humans into means for financial ends).
Grounding ourselves in a common principle like this proposed Zeroth Law of AI Implementation reminds us of the core purpose of technological advancement. It gives us a reference point we can use to evaluate and challenge our decisions, and a shared foundational platform we can build further principles on.
Discuss
Right now, at the start of our future with AI, we have a unique opportunity to talk about where we want to go next and how we want to get there. That conversation starts with talking about core principles. The Zeroth Law of AI Implementation is my contribution to this conversation. I’d love to hear yours!
Cross-posted to LinkedIn.