大規模言語モデルの原理と使いこなしの原則 / Principles of Large Language Models and Their Use

早稲田大学大学院経営管理研究科「プロンプトエンジニアリング ─ 生成AI の応用」2025 春のオンデマンド教材 第4回で使用したスライドです。

Transcript

1. Generated by Stable Diffusion 3.5 — AI 2025 4 (WBS) 2025 4 — 2025-04 – p.1/23
2. https://speakerdeck.com/ks91/collections/prompt-engineering-2025-spring 2025 4 — 2025-04 – p.2/23
3. ( 20 ) 1 • 2 • 3 • 4 • 5 6 RPG 7 “September 12th” 8 9 10 ∼ 11 Linux (Windows )(Mac ) 12 Open Interpreter ∼ 13 01 ∼ 14 AGI (Artificial General Intelligence) 7 (4/28 ) / (2 ) OK / 2025 4 — 2025-04 – p.3/23
4. ( ) ChatGPT ( ) ( ) 2025 4 — 2025-04 – p.4/23
5. ChatGPT ( 4.5 preview) ← ChatGPT – Deep Research ← ( )( 21 ) OpenAI Playground (gpt-4o) ← ( 4.1 ) Google ( ) ← Perplexity ← Grok ← Claude ← o3 o1 GPT-3.5 GPT-4 . . . 2025 4 — 2025-04 – p.5/23
6. o1 2025 4 — 2025-04 – p.6/23
7. o3-mini 2025 4 — 2025-04 – p.7/23
8. Perplexity ( ) ( ) 2025 4 — 2025-04 – p.8/23
9. o1 pro 2025 4 — 2025-04 – p.9/23
10. Perplexity ( ) ( ) 2025 4 — 2025-04 – p.10/23
11. 2025 4 — 2025-04 – p.11/23
12. ChatGPT GPT ChatGPT GPT OpenAI (GPT-3.5, GPT-4{o|.5|.1}, o1, o3) GPT Generative Pre-trained Transformer ( ) (deep learning) a GPT ( ) GPT-3.5, GPT-4 a : ( ) 2025 4 — 2025-04 – p.12/23
13. — ( ) ↓ ELSIE PREPARE TO MEET THY GOD ( ( )) e t h ← Wikipedia “re” (2 ) “e[ ]” (2 ) “th” “the” “th” 1 “th-e” “art-ific-ial”’ 2025 4 — 2025-04 – p.13/23
14. —          (p = 0.xx) (p = 0.yy) . . . GPT ( ) GPT 2025 4 — 2025-04 – p.14/23
15. attention ( ) ( ) GPT / / / / / 2025 4 — 2025-04 – p.15/23
16. ( → → ) OpenAI API "TTJTUBOUT $IBU $PNQMFUJPOT ػೳͷ֊૚ͱͯ͠ݟΔ ग़དྷΔ͜ͱͷ֦͕Γͱͯ͠ݟΔ $PNQMFUJPOT $IBU "TTJTUBOUT ଓ͖Λॻ͍ͯ͘ΕΔ νϟοτ΋Ͱ͖Δ ิ׬͢ΔػೳΛ Ԡ༻͢Δ ର࿩͢ΔػೳΛ Ԡ༻͢Δ ഇࢭ΁ ഇࢭ΁ ͲΜͳ૬खͳͷ͔΋ ͋Β͔͡ΊϓϩάϥϛϯάͰ͖Δ API : Application Programming Interface ( ) 2025 4 — 2025-04 – p.16/23
17. GPT ( ) GPT Alec Radford, Karthik Narasimhan, Tim Salimans, and Ilya Sutskever. 2018. “Improving Language Understanding by Generative Pre-Training”. Available at: https://cdn.openai.com/research-covers/language-unsupervised/language_understanding_paper.pdf. GPT-2 Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. “Language Models are Unsupervised Multitask Learners”. Available at: https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf. GPT-3 Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray, Benjamin Chess, Jack Clark, Christopher Berner, Sam McCandlish, Alec Radford, Ilya Sutskever, and Dario Amodei. 2020. “Language Models are Few-Shot Learners”. Available at: https://doi.org/10.48550/arXiv.2005.14165. GPT-4 OpenAI. 2023. “GPT-4 Technical Report”. Available at: https://doi.org/10.48550/arXiv.2303.08774. 2025 4 — 2025-04 – p.17/23
18. ( = ) ( ) 2025 4 — 2025-04 – p.18/23
19. GPT ( ) GPT Alec Radford, Karthik Narasimhan, Tim Salimans, and Ilya Sutskever. 2018. “Improving Language Understanding by Generative Pre-Training”. Available at: https://cdn.openai.com/research-covers/language-unsupervised/language_understanding_paper.pdf. GPT-2 Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. “Language Models are Unsupervised Multitask Learners”. Available at: https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf. GPT-3 Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray, Benjamin Chess, Jack Clark, Christopher Berner, Sam McCandlish, Alec Radford, Ilya Sutskever, and Dario Amodei. 2020. “Language Models are Few-Shot Learners”. Available at: https://doi.org/10.48550/arXiv.2005.14165. GPT-4 OpenAI. 2023. “GPT-4 Technical Report”. Available at: https://doi.org/10.48550/arXiv.2303.08774. 2025 4 — 2025-04 – p.19/23
20. Generative Pre-Training, Language Models (GPT, GPT-2, GPT-3) : : Improving Language Understanding (GPT : 1.17 ) ( ) ( ) Unsupervised Multitask Learners (GPT-2 : 15 ) Few-Shot Learners (GPT-3 : 1,750 ) → GPT 2025 4 — 2025-04 – p.20/23
21. ( ) ( ← ChatGPT ) ( ) . . . . . . BibTEX ( ) HTML (abstract) PAT ( ) ← ( AI ) AI 2025 4 — 2025-04 – p.21/23
22. ( ) GPT 3 1. (GPT ) 2. ( ) 3. ( ) 3 (↑ 3. ) 3. (GPT ) (↑ 3 ) 2025 4 — 2025-04 – p.22/23
23. 2025 4 — 2025-04 – p.23/23