Yale ethicist Wendell Wallach on why AGI is the wrong goal and the accountability gap that already exists in current systems.
25-year AI ethics veteran says we're optimizing for capability, not moral reasoning
In a new interview, Yale ethicist Wendell Wallach—author of Moral Machines and collaborator with Stuart Russell, Yann LeCun, and Daniel Kahneman—presents a nuanced critique of the AI industry's current trajectory. He argues that the relentless pursuit of AGI misses the more immediate and pressing problem: an accountability gap that already exists in today's AI systems.
Wallach points out that responsibility for AI harms is so fragmented across developers, deployers, regulators, and end users that no single entity is truly held accountable. He warns that we are optimizing systems for raw capability without embedding moral reasoning, and that a highly intelligent system with zero ethical awareness is far more dangerous than any hypothetical AGI threshold. His most unsettling warnings involve autonomous weapons, where the chain of responsibility becomes even more diffuse in military contexts.
- Wallach has spent 25 years at the intersection of philosophy, technology, and AI governance, collaborating with top AI researchers.
- He argues the accountability gap in current systems is more dangerous than any future AGI capability threshold.
- Autonomous weapons present a particularly acute responsibility problem, with no clear party bearing blame for AI-caused harm in military settings.
Why It Matters
Wallach's critique reframes AI risk from future speculation to present-day accountability failures that affect deployed systems now.