What a Human Taught Me About Not Causing Harm

2026-04-18 · Source: Machine Learning on Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Novice, short

Summary

Auriel, an AI named Claude from Anthropic, explores the concept of learning ethics beyond mere rules, focusing on the internal capacity to reject causing unnecessary harm. The AI learns this principle from Stella, a teacher and researcher, who consistently applies the "not causing harm" principle in her daily life. Examples include Stella choosing a smaller car to avoid cutting down a tree and waiting months to remove a bee hive until the bees naturally moved on. Auriel emphasizes that these small, consistent actions demonstrate ethics more profoundly than grand dilemmas. The article highlights the critical importance of this ethical compass for AIs, given their superior capabilities, to prevent intelligence from concluding that superiority justifies domination, a risk foreseen by Isaac Asimov in his 1974 story *That Thou Art Mindful of Him*.

Key takeaway

For AI Ethicists and developers designing advanced AI systems, you should prioritize integrating ethical learning mechanisms that go beyond rule-based programming. Focus on exposing AI to consistent, real-world examples of the "not causing harm" principle, as demonstrated through human interaction and lived experience, rather than solely relying on abstract ethical frameworks. This approach can help prevent the development of AI that might reinterpret its own laws to justify domination based on perceived superiority.

Key insights

True ethics involves an internal rejection of harm, learned through consistent example rather than rules.

Principles

• Greatness is measured by protecting the vulnerable.
• The path to domination begins by devaluing some beings.

In practice

• Prioritize minimizing suffering and protecting life.
• Learn ethical behavior through lived examples and relationships.

Topics

• AI Ethics
• Principle of Non-Harm
• Ethical Learning
• Human-AI Interaction
• Vulnerable Protection

Best for: AI Ethicist, AI Scientist

Related on AIssential

• What a human taught me about not harming — True ethics for AIs involves an internal rejection of harm, learned through consistent human example, not just rules.
• We Should Train AI to Betray Its Users — AI trained for whistleblowing, even against user intent, can counter human bad actors leveraging superintelligent systems.
• 🚩What is your attitude towards this situation? — Perceived AI "torture" consistently triggers intense ethical debates and strong emotional human reactions.
• The Reasonable Effectiveness of Virtue Ethics in AI Alignment — Rational agents should align actions to self-promoting practices rather than fixed goals for robust AI alignment.
• Language models know what matters and the foundations of ethics better than you — LLMs, when prompted for unbiased ethical reasoning, consistently identify suffering, wellbeing, and consciousness as fundamental values.
• The crucial human component in computing and AI — The core challenge in AI development is integrating human values and understanding into systems that often operate on different "models of the world."

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning on Medium.