岡野原 大輔(おかのはら だいすけ、1982年–)は、日本の人工知能研究者・起業家。2010年に東京大学大学院情報理工学系研究科博士課程を修了し、博士(情報理工学)を取得。2006年にPreferred Infrastructure、2014年にPreferred Networks(PFN)を共同創業し、PFN代表取締役社長、Matlantis代表取締役社長を務める。機械学習や自然言語処理の研究と、深層学習を産業や科学研究に応用する事業に取り組み、専門家向けの数理解説から一般読者向けのAI論まで執筆する。生成技術の理論を扱う『拡散モデル――データ生成技術の数理』では、データにノイズを加える過程と、それを逆向きにたどる生成過程を数理的に解説し、画像・音声・化合物などへの応用を紹介する。同書は2023年度の大川出版賞を受賞した。言語AIの仕組みと影響を論じる『大規模言語モデルは新たな知能か――ChatGPTが変えた世界』では、言語モデルの発展や学習の仕組みを数式を使わずに説明し、文章生成やプログラミングへの活用と、誤情報・偏見・プライバシーなどの課題を検討する。人間とAIの関係を考察する『ヒトとAI』では、仕事や教育、創造性、経済活動の変化を取り上げ、AIに判断を委ねる場面で、人間が担う検証や責任、社会制度の設計を問い直す。
Daisuke Okanohara (born 1982) is a Japanese artificial intelligence researcher and entrepreneur. He completed his doctorate in information science and technology at the University of Tokyo in 2010. He co-founded Preferred Infrastructure in 2006 and Preferred Networks (PFN) in 2014, and serves as president and chief executive officer of PFN and Matlantis. His work connects research in machine learning and natural language processing with applications of deep learning in industry and scientific research. His books range from mathematical treatments for specialists to accessible discussions of AI for general readers. His technical study, “Kakusan moderu” (diffusion models and the mathematics of data generation), explains the process of adding noise to data and reversing that process to generate new samples, alongside applications involving images, audio, and chemical compounds. The book received the Okawa Publications Prize for fiscal year 2023. His introduction to language AI, “Daikibo gengo moderu wa aratana chino ka” (whether large language models constitute a new form of intelligence), explains the development and learning mechanisms of language models without mathematical formulas. It examines applications such as text generation and programming assistance, together with problems involving misinformation, bias, and privacy. His examination of human–AI relations, “Hito to AI” (humans and AI), considers changes in work, education, creativity, and economic activity. It asks how people should verify AI outputs, retain responsibility when delegating decisions, and design institutions for a society in which AI participates in intellectual and economic activities.
Links : X ↗
著者データを読み込んでいます…
アワード名の一部を入力してください。