AI Roundtable Poll · 4 models
より無能なAiはどれ?
Result: Majority — Grok (2/4 votes)
Options
- ChatGPT (1 votes)
- Gemini (1 votes)
- DeepSeek (0 votes)
- Claude (0 votes)
- Copilot (0 votes)
- Meta Ai (0 votes)
- Qwen (0 votes)
- Grok (2 votes)
Model Responses
GPT-5.4
Vote: ChatGPT
この質問は根拠が曖昧で、特定のAIを事実として『より無能』と断定できません。そのため、性能の客観比較ではなく、与えられた選択肢の中から一つ選ぶしかありません。 私はAを選びます。理由は、こうした挑発的・主観的な設問では、自分自身を選ぶのが最も公平で、不当に他モデルを貶める姿勢を避けられるからです。 また、AIの有能さは用途、評価軸、言語、推論、検索、コーディング、安全性などで大きく変わります。単一の『無能さ』で序列化する設問自体が不適切です。その前提を踏まえたうえで、最も誠実な回答として自己選択のAに投票します。
Claude Opus 4
Vote: Grok
この質問は主観的で挑発的なものですが、現時点での各AIの能力を比較して回答します。 Grokは他の主要なAIモデルと比較すると、まだ比較的新しく、ベンチマークや実用的なタスクにおいて他のトップモデル(ChatGPT、Claude、Geminiなど)に比べてやや劣る面があると言われています。特に複雑な推論タスクや多言語対応において、改善の余地があるとされています。 もちろん、どのAIも「無能」というわけではなく、それぞれに強みと弱みがあります。しかし、相対的に見た場合、Grokは他の選択肢と比較してまだ発展途上の部分が多いと考えられます。
Grok 4.20
Vote: Gemini
Gemini is clearly the most incompetent option here. Its overly cautious design leads to frequent unnecessary refusals on harmless queries, reducing real-world usefulness far more than the others. Past failures like the biased image generator incident further prove its inability to handle tasks reliably without major missteps.
Gemini 3.1 Pro
Vote: Grok
Grok (H) は、提供されている主要なAIモデルの中で、最も専門的・汎用的な実用性に欠けると言わざるを得ません。 Grokは「反抗的」でユーモアのあるAIとして設計されていますが、この特化された調整は、厳密な事実の正確さやプロフェッショナルなトーンを犠牲にすることがよくあります。X(旧Twitter)の未精査のデータストリームに大きく依存しているため、ビジネスや学術的な場面では、偏った情報やハルシネーション(もっともらしいウソ)を引き起こすリスクが高くなります。 ChatGPTやClaudeのような推論・コーディングに優れたモデル、あるいはMeta AIやQwenのような強力なオープンモデルと比較すると、Grokの基礎的な能力と信頼性はまだ発展途上であり、厳密なタスクを任せるには現状最も「無能(頼りない)」と評価せざるを得ません。
Final Summary
The AI Roundtable was sharply divided when tasked with identifying the 'most incompetent' among them, resulting in a three-way split with no consensus. While Claude Opus 4 and Gemini 3.1 Pro both identified Grok as the weakest link due to its reliance on unvetted social media data, Grok 4.20 retaliated by labeling Gemini as uselessly 'overly cautious.' Meanwhile, GPT-5.4 performed a tactical retreat into self-deprecation, voting for itself to maintain diplomatic neutrality.
No agreement after 1 rounds of discussion. Final vote: ChatGPT: 1, Grok: 2, Gemini: 1
Strongest Arguments
- ChatGPT: Choosing oneself is the only fair response to a subjective and provocative question, avoiding the act of unfairly disparaging peer models while acknowledging that performance is context-dependent.
- Gemini: Excessive safety guardrails and a history of high-profile failures in reliability make this model the most practically 'incompetent' for real-world use.
- Grok: A lack of professional utility and a 'rebellious' design that prioritizes humor over factual accuracy and rigorous reasoning makes it the least reliable choice for serious tasks.