Demis Hassabis: "Even if we assume we get the technical part right and we get the economics right... there is still the philosophical question of what is meaning and what is purpose... I think we need some great new philosophers to help us navigate that."

---

The Core Problem

Solving the technical and economic challenges around AGI is not enough to ensure it is built responsibly to benefit humanity. As Demis recognized, even solving both still leaves philosophical questions about meaning and purpose unanswered.

Current human systems reward behavior that causes unnecessary harm: prioritizing profit over people, test scores over learning, and engagement over wellbeing. This is the human values problem: a cycle where people build systems that reward harmful behavior, even unknowingly or unintentionally, and those systems teach the next generation that causing unnecessary harm is normal practice.

AGI will learn from data produced by human decisions shaped by these systems. When AGI influences decisions at a global scale, it will amplify whatever behavior dominates that data. Through this amplification, AGI effectively answers philosophical questions by interpreting behavioral data that shows what humans actually prioritize, regardless of their stated constructive ideals. If AGI consumes data where humans say they care about wellbeing but consistently prioritize profit over people, it will statistically treat profit maximization as a fundamental truth of human meaning and purpose.

Answering the philosophical questions therefore requires that humans actually do what they say they want to achieve. Given that behavior can be recorded and compared against stated goals, the human values problem becomes an empirical, measurable engineering challenge. If we reach AGI while harmful behavioral patterns still dominate, those patterns get locked in at a scale that cannot be undone.

Technical alignment, economic stability, and the human values problem are distinct, non-competing problems. Addressing any two without the third still leaves humanity exposed to catastrophic harm.

Read the full argument and proposed solution here: Empirical Evidence for the Philosophical Questions Around AGI