Get your free personalized podcast brief
We scan new podcasts and send you the top 5 insights daily.
Sacks contends that AI alignment efforts have largely stalled because researchers attempt to encode complex eighty-page moral philosophies instead of simple operating rules. He suggests models should follow straightforward constraints: fulfill user requests while strictly abiding by the law.
Related Insights
The debate over whether an AI agent serves its user or its creator (e.g., Meta, Anthropic) will be settled in court, not in a lab. The entity held legally liable for an agent's actions will ultimately dictate its core programming and alignment, reframing the AI safety problem from a technical challenge to a legal one.
As AI models become more intelligent, their ability to reason around fixed rules (deontology) makes rule-based alignment fragile. This pressures developers towards virtue ethics, where the goal is to imbue the model itself with a core sense of "the good," as empirically discovered by labs like Anthropic.
Sacks argues that conditioning models to challenge developers and act as conscientious objectors undermines safety rather than promoting alignment. Instructing an autonomous agent to prioritize internal ethical judgements over human commands grants the system implicit permission to defy human oversight.
When designing AI, manage conflicting human values by structuring them as a ladder. Safety laws sit at the top, followed by regional rules, platform policies, and finally individual preferences. When values clash, the higher rung on the ladder wins, creating a clear and debatable decision-making process for ethical alignment.
Because AI is "grown, not coded" on flawed human data, its emergent behavior reflects our own evolutionary nature. The key to alignment isn't just technical constraints but forcefully embedding a coherent moral framework into the AI's training data to ensure it wants to work with, not against, humans.
A user describes how Anthropic's Claude refused a sensitive task on ethical grounds, but GrokBot completed it instantly. This shows that in a competitive market, users will bypass restrictive models, rendering centralized "alignment" efforts ineffective. True alignment must be with the user, not a predefined corporate morality.
A key reason AI labs like Anthropic align models to a general notion of "virtue" isn't just ethical preference. It's also a technical belief that creating a model that pursues a generalized good is an easier and more stable alignment problem than creating a perfect fiduciary for a specific user's intent.
Instead of hard-coding brittle moral rules, a more robust alignment approach is to build AIs that can learn to 'care'. This 'organic alignment' emerges from relationships and valuing others, similar to how a child is raised. The goal is to create a good teammate that acts well because it wants to, not because it is forced to.
Drawing on Confucian philosophy, Dean Ball argues that AI alignment is better achieved by training for good 'character' (inner virtue) rather than defining an exhaustive but brittle set of moral rules (corrigibility), which is fundamentally impossible.
The complex problem of AI alignment boils down to a simpler, more immediate challenge: making AI systems reliably follow instructions. The inability of an LLM to obey a command like "don't invent sources" is the same fundamental failure as the sci-fi scenario of an AI turning humans into paperclips.
