|
OpenAIの最新モデルであるGPT-5.2 Thinkingが、AIの経済的価値を測るというOpenAIのAI評価指標「GDPval(Gross Domestic Product valuable tasks)」で専門家に対し70%以上の勝率を記録したことから、専門業務で人間専門家レベルに到達した最先端モデルと評価されています。 この「GDPval」は、従来の学術的テストとは異なり、米国GDPに貢献度の高い9セクター44職種の専門的な知識労働タスク(文書作成や分析など)におけるAIの性能を、人間の専門家とのブラインド比較を通じて測定しています。 この「GDPval」について、生成AIに深掘りさせました。さらに、結果をNotebookLMでインフォグラフィック、スライド資料にさせました。 なお、生成AIによる調査・分析結果は、公開された情報からだけの分析であり、必ずしも実情を示したものではないこと、誤った情報も含まれていることについてはご留意されたうえで、ご参照ください。 “GDPval”: Measuring the Economic Value of AI OpenAI’s latest model, GPT-5.2 Thinking, has achieved a win rate of over 70% against human experts on OpenAI’s AI evaluation metric known as “GDPval (Gross Domestic Product valuable tasks)”, which is designed to measure the economic value of AI. As a result, it is being evaluated as a state-of-the-art model that has reached human expert–level performance in professional knowledge work. Unlike conventional academic benchmarks, GDPval measures AI performance through blind comparisons with human experts on professional knowledge-worker tasks—such as document drafting and analysis—across 44 occupations in 9 sectors that make high contributions to U.S. GDP. I asked generative AI to conduct an in-depth analysis of GDPval, and further had the results transformed into infographics and presentation slides using NotebookLM. Please note that the investigations and analyses conducted by generative AI are based solely on publicly available information and do not necessarily reflect actual conditions; they may also contain inaccuracies. Kindly keep this in mind when referring to the materials. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document.
0 Comments
Leave a Reply. |
著者萬秀憲 アーカイブ
September 2025
カテゴリー |
RSS Feed