A Philosophical CEO of Anthropic Needs to Study Some Economics

When the CEO of Anthropic was young, he was thinking about society’s Social Welfare Function.

He was saying that he would be willing to sacrifice a human if this action would save X chickens (perhaps X=million, a billion, a trillion).

He has the right to have that pro-animal worldview. Similar to a Samuel Jackson in the movie Pulp Fiction, he is meting out justice in a righteous way. But, he is ignoring the power of the Coase Theorem! Here is thread I posted.

We all did and said silly stuff when we were young but does the CEO of this Trillion dollar company now embrace market approaches for solving problems?

I posed my entry to Chat GPT. Here is what Chat GPT wrote.

CHAT GPT’s Response

Your Coase intuition is useful, but you claim more than it establishes. The strongest point is that a dramatic moral dilemma may conceal peaceful alternatives. The weak point is claiming that those alternatives necessarily resolve the dilemma—or refute Amodei’s ethical position.

I would tighten five parts of the argument:

1. Distinguish a welfare ranking from a policy recommendation. In the excerpt, Amodei accepts that sufficiently many chicken lives could outweigh one human life in a utilitarian calculation. Given the question, that has an unsettling implication. But “condoned murder” implies a broader endorsement than this recalled answer establishes. Your Pulp Fiction comparison also attributes righteous aggression that the excerpt does not demonstrate.

2. Coase expands the feasible choices; he does not settle the ethical ranking. Your question—“Why not pay this person to stop?”—is valuable. But someone could agree with your proposed bargain while retaining Amodei’s ranking for circumstances where bargaining fails. You have challenged the assumption that killing is necessary, rather than disproved the underlying social welfare function.

3. Bargaining does not guarantee that every chicken survives. A voluntary agreement requires the rescuers’ willingness and ability to pay to exceed the killer’s minimum acceptable compensation. Even under ideal bargaining conditions, efficiency does not necessarily mean zero chicken deaths. Coase’s broader contribution also requires attention to the costs of organizing, contracting, and enforcing agreements. sites.duke.edu Your “club” therefore needs an explanation of fundraising, free riding, and enforcement.

4. Your second case is principally about enforcing rights. If chickens already possess an enforceable right to live, appointing guardians or funding their protection could help implement that right. Adoption is one possible arrangement, but private ownership alone does not establish protection: the owner might want to kill the chicken. The crucial institution is an enforceable restriction on killing, together with someone empowered to defend it.

5. Paying people to stop threatening animals creates a dynamic incentive problem. A payment might save today’s chickens while encouraging tomorrow’s threats. A stronger economic argument considers both the immediate bargain and the behavior that a standing rescue policy would induce.

The most interesting version of your piece asks: “Before calculating whom to sacrifice, have we exhausted the opportunities for mutually beneficial exchange?” That is a sharp criticism of thought experiments that artificially restrict the choice set.

I would replace “Coase offered a far cheaper way to solve the problem” with “Coase prompts us to ask whether bargaining could eliminate the supposed need for sacrifice.” Then connect it to AI leadership: How much effort should powerful decision makers devote to finding voluntary solutions before imposing their preferred welfare tradeoffs?

That gives you a substantive institutional argument. The childhood anecdote alone cannot establish his current attitude toward markets.