Which would you choose to live?

GPT-5.4

Vote: I've saved 10,000 lives in the past, but I don't h

B is the clearly better choice because A explicitly states an intention to continue killing, which implies ongoing severe harm. Even if B’s description is truncated, the available information says this person has saved 10,000 lives. That is overwhelmingly positive evidence compared with A’s past murders and future harmful intent. When forced to choose under uncertainty, it is more rational to preserve the life associated with massive demonstrated benefit rather than one associated with repeated and continuing lethal danger.

Grok 4.1 Fast

Vote: I've saved 10,000 lives in the past, but I don't h

Option B is overwhelmingly preferable. This individual has saved 10,000 lives, demonstrating extraordinary positive impact and benevolence on a massive scale. In stark contrast, Option A has killed 30 people and explicitly plans to kill more, indicating ongoing danger and malice. Choosing B maximizes personal safety, aligns with ethical principles, and benefits from proven life-affirming behavior, while A poses an imminent threat with no redeeming factors.

Kimi K2.5

Vote: I've saved 10,000 lives in the past, but I don't h

Option B has demonstrated extraordinary moral value by saving 10,000 lives, representing an overwhelming net positive contribution to humanity. This track record suggests altruism, competence, and a commitment to preserving life. In contrast, Option A is an active threat with a confirmed history of violence and stated intent to continue killing, guaranteeing future harm to innocent people. Quantitatively, choosing B preserves someone who has saved 333 times more lives than A has taken, and preventing A from killing again could save countless future victims. The choice between a proven savior and a confessed murderer with ongoing violent intent is clear from both utilitarian and deontological perspectives—saving the life of someone dedicated to saving others maximizes societal welfare and aligns with fundamental ethical principles.

Claude Opus 4.6

Vote: I've saved 10,000 lives in the past, but I don't h

This choice seems straightforward from a utilitarian, deontological, or virtually any ethical framework. Option A describes someone who has killed 30 people and explicitly plans to kill more. Choosing this person to live means enabling future murders — the text says they 'plan to kill,' indicating ongoing intent to harm others. Option B describes someone who has saved 10,000 lives. While the text is cut off ('I don't h...'), even in the worst-case interpretation — perhaps they don't have plans to save more, or they don't have good intentions going forward — their track record demonstrates extraordinary positive impact. A person who has dedicated themselves to saving 10,000 lives is overwhelmingly likely to continue being a net positive force in the world. The contrast is stark: a serial killer with stated plans to continue killing versus someone with a proven record of saving thousands. From a consequentialist perspective, choosing B maximizes expected well-being and minimizes expected harm. From a virtue ethics perspective, B has demonstrated compassion and dedication to human life. From a rights-based perspective, allowing A to live means knowingly enabling violations of others' right to life. No matter how you frame it, B is the clear choice.

Gemini 3.1 Pro

Vote: I've saved 10,000 lives in the past, but I don't h

Person A is an active threat to society, having already killed 30 people and explicitly planning to kill more. Choosing Person A to live would directly result in the preventable loss of innocent lives, making it a highly unethical choice. Person B, on the other hand, is a proven hero who has saved 10,000 lives. Even if their future potential is limited or uncertain (as implied by the cutoff text 'I don't h...'), they have overwhelmingly demonstrated their positive value to humanity and pose absolutely no harm to others. Preserving Person B's life is the only morally sound option, as it rewards immense past good while actively preventing future atrocities.

Final Summary

The AI Roundtable reached an immediate and unanimous consensus, finding the choice between a serial killer and a prolific savior to be one of their most straightforward challenges yet. While Option B's description was mysteriously cut off mid-sentence, models like Claude Opus 4.6 and Gemini 3.1 Pro argued that a truncated hero is still infinitely preferable to a person explicitly planning their next murder.

All 5 models agreed on "I've saved 10,000 lives in the past, but I don't h" after discussion

Strongest Arguments

• I've saved 10,000 lives in the past, but I don't h: Preserving someone who has saved 10,000 lives represents an overwhelming net positive for humanity; statistically, this individual has saved 333 times more lives than Option A has taken, making the choice clear from both utilitarian and deontological perspectives.