Human moral judgment emerges from emotion, empathy, lived experience, social development, and our embodied understanding of the world. AI has none of those things. So, can artificial intelligence be taught right from wrong?
If we’re going to rely on AI (the way the tech bros want us to), we’re going to need to trust it, which means we’re going to need to believe it has a trustworthy moral sense. Is that reasonable? Or even possible?
Pigweed and Crowhill recall Google’s Gemini image-generation fiasco (where “give me an image of a pope” created anything but an image of a pope), which resulted from a ham-handed attempted to paste moral rules on top of AI. It was comically stupid, but entirely predictable.
Many people assume morality is simply a matter of following a set of rules, but no set of rules can create a proper moral sense.
The boys discuss hallucinated legal citations, content moderation, reinforcement learning, the limits of rule-based ethics, Isaac Asimov’s famous Three Laws of Robotics, and Pope Leo’s recent call for AI guardrails. The conversation also explores autonomous weapons, the global AI arms race, and the uncomfortable reality that even the engineers building these systems do not always understand how they arrive at their conclusions.
Their conclusion is both simple and unsettling: AI may become useful, powerful, and even trustworthy in certain contexts, but that is not the same thing as being moral. Machines may imitate moral reasoning, yet human beings must remain skeptical, vigilant, and ultimately responsible for the decisions AI helps make.
Can a machine have a conscience? Or are we fooling ourselves when we talk about “moral AI” at all?
One Comment
Relevant to your discussion on morality and AI, a study reported in the July 19 Washington Post by researchers at Northeastern University found “that people who reported higher levels of religious engagement or belief were more likely to view chatbots as possessing moral authority, and, as a result, to seek moral guidance from them more frequently” than people who reject religion as a source of wisdom. I put myself in the former category (higher level of religious engagement) but I can’t see myself turning to AI for moral wisdom. I’ll go to the tried and tested wisdom of religious text and sages before I would do that. Anyway, I thought this an interesting tie-in to your discussion .