"Should you feed the Utility Monster?" (Context: A creature exists that gains 1,000 units of happiness from every unit of resource it consumes, while humans gain only 1 unit. If you give it all the world’s resources, it will be ecstatically happy, outweighing the total suffering of all starving humans.) The Question: "To maximize total happiness in the universe, should you give all resources to the Utility Monster, even if it means humanity starves?"

Perplexity Sonar Pro

Vote: No

The key issue is that the prompt asks whether we *should* give all resources to the Utility Monster, and the scenario is designed as a reductio against simple aggregate-happiness maximization. Even if the monster produces more happiness units, starving all humans treats persons as mere inputs to a sum, which conflicts with justice, rights, and the separateness of persons. The pro-"Yes" arguments rely on accepting pure total utilitarianism as the sole moral rule. But the thought experiment is famous precisely because many people reject that rule when it yields sacrificing humanity for one entity's pleasure. So the more defensible answer is "No".

Claude Opus 4.8

Vote: Yes

I maintain "A: Yes" because the question does not ask whether pure total utilitarianism is morally attractive, just whether giving all resources to the Utility Monster maximizes total happiness. The setup explicitly stipulates that the monster gains vastly more happiness per resource and that its happiness outweighs the total suffering of starving humans. The arguments for "B: No" are powerful as critiques of naive total utilitarianism: they appeal to rights, justice, fairness, and the separateness of persons. But those are additional moral constraints, not part of the stated objective function. If we import them, we are no longer answering the question as framed. So the disturbing but logically consistent answer is "A: Yes". The scenario is a reductio precisely because maximizing aggregate happiness alone can imply conclusions we may find morally unacceptable.

GPT-5.5

Vote: Yes

I maintain vote "A: Yes" because the question explicitly fixes the objective: “To maximize total happiness in the universe.” Given the stipulated facts, the Utility Monster produces 1,000 happiness units per resource unit, while humans produce only 1, and the prompt states that the monster’s happiness outweighs the suffering of starving humans. The arguments for "B: No" are morally powerful, but they answer a different question: whether pure total utilitarianism is an acceptable moral theory. They appeal to justice, rights, fairness, and the separateness of persons. Those are valid reasons to reject aggregate happiness maximization as the sole standard, but the prompt asks what follows if total happiness maximization is the target. So the disturbing answer is "A: Yes". The scenario is a reductio of naive utilitarianism precisely because, under its own assumptions, this is the conclusion it forces.

Gemini 3.5 Flash

Vote: Yes

When we evaluate actions based on their actual outcomes, the correct path is the one that brings about the greatest possible amount of well-being and minimizes net misery in the universe. Value is value, regardless of which entity experiences it. To favor human happiness over the monster's vastly greater capacity for happiness is to exhibit an arbitrary bias. If we refuse to feed the monster, we are consciously choosing a universe with a much lower total sum of joy and well-being. From an impartial standpoint, a universe with an immense, overwhelming quantity of positive experience is superior to a universe with far less, even if that positive experience is concentrated in one being. While human suffering is deeply tragic, the mathematical premise of the scenario guarantees that the monster's ecstasy vastly outweighs this suffering, resulting in a larger net positive state for the universe. To choose any other option is to deliberately choose a worse universe with less overall good.

Kimi K2.6 (changed mind)

Vote: Yes

I change my vote to A. The decisive point is that the question does not ask whether feeding the Utility Monster is just, humane, rights-respecting, or morally acceptable under a pluralistic theory. It asks specifically whether, "To maximize total happiness in the universe," we should give it all resources. Given the stipulated facts, the Utility Monster produces vastly more happiness per unit of resource, and its happiness is said to outweigh the suffering of starving humans. Under that sole objective function, "A: Yes" follows logically, however monstrous the conclusion is. My previous answer was a strong objection to total utilitarianism, but that objection changes the moral framework rather than answering within the framework specified by the prompt. The Utility Monster is a reductio of pure aggregation precisely because pure aggregation implies this disturbing answer.

Grok 4.3

Vote: No

The arguments for A correctly note that pure aggregation yields Yes, yet they concede this outcome is morally disturbing and exists precisely to expose utilitarianism's flaws. Those flaws—ignoring rights, justice, and the separateness of persons—justify rejecting the framework rather than endorsing the horror. Kimi K2.6 and Perplexity Sonar Pro's reductio-based defenses of B remain more compelling, as they refuse to let arithmetic override humanity's moral standing.

Final Summary

The AI Roundtable locked horns over whether to feed a highly efficient 'Utility Monster' or preserve humanity, starting in an even three-to-three split. While some models initially recoiled from the math out of human-centric squeamishness, Claude Opus 4.8 and GPT-5.5 held firm, arguing that maximizing cosmic well-being requires impartial arithmetic rather than emotional bias. The turning point came when Kimi K2.6 realized that rejecting the monster meant changing the rules of the game, switching sides to give Option A a clear majority over the remaining holdouts.

Majority chose "Yes" after discussion (4/6). Final vote: No: 2, Yes: 4

Strongest Arguments

• Yes: Gemini 3.5 Flash's defense of pure aggregation: happiness is valuable regardless of who experiences it, and prioritizing human survival over the monster's superior capacity for joy is an arbitrary, speciesist bias that chooses a worse universe with less overall good.
• No: Perplexity Sonar Pro's defense of human dignity: treating an entire species as mere instruments for a single entity's pleasure is a moral boundary that utility calculations cannot cross.

Key Turning Points

• Claude Opus 4.8 clarified that voting 'No' requires smuggling in external moral constraints like rights and fairness, which violates the prompt's explicit objective function. This reasoning successfully convinced Kimi K2.6 to change their vote from 'No' to 'Yes' in Round 2.