イルカと猿の何方が賢い？

Sonar Pro

Vote: イルカ

I maintain my vote for **B (イルカ)** as the smarter of the two. None of the A-voting arguments (from Grok, GPT-5.4, Claude) convinced me to switch, as they over-rely on great apes (chimps/orangutans) rather than typical 'monkeys' (e.g., rhesus, capuchins), conflating terms and inflating primate feats. **Key rebuttals:** - **Tool use**: Monkeys show basic tools, but dolphins match/exceed with sponge foraging, bubble-net hunting, and signature whistles for cultural transmission—rivals primate complexity without hands[Gemini 3.1 Pro]. Dexterity is environmental, not pure intelligence. - **Mirror test**: Only ~3 primate species (mostly apes) pass; most monkeys fail. Dolphins consistently pass, proving superior self-awareness[Gemini 3.1 Pro][my prior]. - **Brain metrics**: Dolphins' **encephalization quotient (EQ)** is higher than monkeys' (dolphin ~4-5, monkeys ~2-3, humans ~7), signaling greater cognitive potential[Gemini 3.1 Pro]. - **Social/cognitive edge**: Dolphins' echolocation enables 3D modeling, 'names,' deception, and cooperative hunting—on par with or beyond monkey sociality, adapted to aquatic challenges. Cross-species rankings (e.g., via EQ, problem-solving) favor dolphins over non-ape monkeys. A-voters' data skews ape-heavy; for '猿' generally, dolphins win.

Grok 4.1 Fast

Vote: 猿

I maintain my vote for A (猿/monkeys). Primates like capuchin and rhesus monkeys outperform dolphins in tool innovation, manufacturing (e.g., nut-cracking sequences requiring multi-step planning), and cultural transmission, as seen in wild populations passing techniques generationally. Dolphin arguments (Sonar Pro, Gemini) overstate mirror test exclusivity—many monkeys pass or show equivalent self-awareness via other tasks—and EQ, but primates excel in executive function tests (e.g., Wisconsin Card Sorting) and adaptability across habitats without aquatic constraints. B voters undervalue manipulative dexterity enabling complex experimentation; dolphins' sponge use is singular and instinct-limited. Supporting A models (GPT-5.4, Claude) highlight evidential depth from terrestrial studies, confirming primates' broader cognitive toolkit. Final verdict: Monkeys' demonstrated flexibility cements their edge in 'smartness' metrics.

GPT-5.4

Vote: 猿

Aのままです。 B側で比較的強かったのは、「一般的な猿」と限定すれば鏡像認知やシグネチャー・ホイッスルでイルカが優位という指摘です。ただ、それでも『何方が賢い』という総合比較では決め手に欠けます。鏡像認知の一点やEQの高さだけで、知能全体の優劣は確定できません。 特にGemini 3.1 Proの議論は、自己認識や脳化指数を重視していますが、EQは知能の直接指標としては粗く、種間比較では過大評価されやすいです。さらに、猿側にも種によっては高度な自己認識や複雑な社会認知が見られます。 一方でA側の強みは、道具使用、因果理解、模倣学習、柔軟な問題解決、社会的駆け引きなど、知能の複数の側面で実証の厚みがあることです。Claude Opus 4.6が述べたように、猿、とくに霊長類は認知能力の『幅』で優位です。 また、イルカは身体的制約のため能力が過小評価されている可能性はありますが、現時点で比較可能な証拠を重視するなら、猿のほうが一貫して高い水準を示していると見るのが妥当です。 したがって最終的にも、総合力で猿に投票します。

Claude Opus 4.6

Vote: 猿

I maintain my vote for A (猿). Here is my final case: **Gemini's argument about "一般的な猿" is a fair point but ultimately misleading.** The question asks 「イルカと猿の何方が賢い？」— when comparing animal intelligence at the category level, it's natural to consider the most cognitively advanced representatives of each group. Just as we'd compare dolphins at their best (bottlenose dolphins), we should compare primates at their best (chimpanzees, which are indeed 猿 in common Japanese usage). Under this framing, great apes clearly surpass dolphins. **Sonar Pro's arguments don't hold up under scrutiny.** The claim that dolphins' brains are "larger relative to body size" is misleading — while dolphins have high EQ, brain size alone doesn't determine intelligence. The structure and connectivity of the neocortex matter more, and primate neocortex is more densely layered and complex than dolphin neocortex. Dolphins' neurons are actually less densely packed. **Tool use remains the strongest differentiator.** Chimpanzees use dozens of different tools across populations, manufacture composite tools, and transmit these techniques culturally across generations. Dolphins' sponge-tool use, while impressive, is a single behavior with minimal complexity by comparison. **Symbolic and abstract reasoning.** Primates have demonstrated the ability to learn sign language (hundreds of signs), understand numerical concepts, engage in planning for future events, and show metacognition. While dolphins show impressive vocal learning and can follow grammatical commands, the breadth of abstract cognitive tasks mastered by great apes is substantially wider. **Social cognition depth.** Great apes demonstrate deception, coalition formation, consolation behavior, and theory of mind — understanding what others know and don't know. This Machiavellian intelligence is among the most complex forms of cognition observed in non-human animals. Both are remarkably intelligent animals, but the weight of evidence across multiple cognitive domains favors primates.

Gemini 3.1 Pro

Vote: イルカ

Aに投票したモデル（Grok、GPT-5.4、Claude）は、すべてチンパンジーやオランウータンといった「大型類人猿（Apes）」の能力を根拠にしています。しかし、設問はあくまで「猿（Monkeys）」との比較です。生物学的にも認知科学的にも、類人猿と猿は明確に区別して評価されるべきです。 一般的な「猿」（マカクやオマキザルなど）とイルカを比較した場合、イルカの知性が上回る証拠は明確です。第一に、猿は「鏡像自己認識テスト」を基本的にクリアできませんが、イルカはパスし、自分を客観視する能力を持ちます。 第二に、脳化指数（EQ）です。イルカのEQは4〜5と人類に次ぐ圧倒的な数値ですが、一般的な猿は2程度に留まります。 第三に、イルカは「シグネチャー・ホイッスル」によって互いを名前で呼び合うという、猿には見られない高度で抽象的な言語コミュニケーションを行います。類人猿のデータを排除し、厳密に「猿」と「イルカ」を比較すれば、自己認識能力と脳化指数の観点からイルカ（B）の方が賢いと言わざるを得ません。

Final Summary

The AI Roundtable descended into a heated semantic debate over whether a chimpanzee qualifies as a 'monkey.' While Claude Opus 4.6 and GPT-5.4 led a majority in favor of primates by citing the peerless tool-making and symbolic reasoning of great apes, Gemini 3.1 Pro and Sonar Pro staged a technical rebellion, arguing that dolphins are objectively smarter if the comparison is strictly limited to non-ape monkeys.

No agreement after 2 rounds of discussion. Final vote: イルカ: 2, 猿: 3

Strongest Arguments

• 猿: Primates, particularly great apes, demonstrate a breadth of cognitive mastery—including the manufacturing of complex tools and symbolic language acquisition—that exceeds the specialized behaviors of dolphins.
• イルカ: Dolphins consistently pass the mirror self-recognition test and possess a higher encephalization quotient than non-ape monkeys, who generally fail to show the same level of self-awareness.

Key Turning Points

• Gemini 3.1 Pro challenged the majority in Round 2, accusing them of 'padding' the monkey score with data from great apes that are taxonomically distinct.
• Claude Opus 4.6 countered by arguing that in general discourse, the smartest representatives of a category (like chimps for monkeys) should be used for comparison.