목차
Title page 1
Contents 1
Abstract 2
1. Introduction 3
2. What is Generative AI? 7
3. Is GenAI a General Purpose Technology? 10
3.1. Diffusion 13
3.2. Knock-on Innovation 17
3.2.1. Products 18
3.2.2. Production processes 19
3.2.3. Organization 20
3.2.4. Case study evidence 20
3.3. Ongoing Core Innovation 21
3.3.1. Model Development 22
3.3.2. Hardware 28
3.3.3. Datasets 33
3.4. The Case that GenAI is a GPT 34
4. Is GenAI an Invention of a Method of Invention? 37
4.1. Indicators of GenAI Research and of GenAI Use in Research 43
4.2. Is GenAI an IMI? 48
5. Conclusion 49
References 52
Appendix 73
A. Definitions of AI 73
B. A Short History of AI 78
Tables 5
Table 1. Selected GenAI Productivity Field Studies 5
Table 2. Examples of General Purpose Technologies 11
Table 3. Price of Compute, Selected Nvidia GPUs 29
Table 4. Examples of Inventions of Methods of Invention 38
Table 5. The Stages of a Research Project 40
Table 6. Scientific Activity beyond Research Projects 42
Table 7. Occupations with High GenAI Usage 45
Table 8. O*NET Scientific Task Prevalence in Claude AI Prompts 46
Table 9. Distinguishing Features of AI Models 78
Table 10. History of Amazon Demand Forecasting Techniques 81
Figures 4
Figure 1. AI Benchmark Performance 4
Figure 2. Indicators of Interest in GenAI 6
Figure 3. Indicators of U.S. AI-Related Investment 7
Figure 4. AI Use Over Time 15
Figure 5. AI-Related Job Postings 16
Figure 6. Number of Parameters and Training Dataset Size 21
Figure 7. Open-Source AI Models 22
Figure 8. Price Indices of GPU Improvements 30
Figure 9. Progress on Moore's Law: Two Perspectives 32
Figure 10. Indicators of Power Demand and Supply 32
Figure 11. AI Mentions in Scientific Patents 44
Figure 12. GenAI Automation vs. Augmentation in Researcher Roles 47
Figure 13. Mentions of AI Usage for Research in Conference Calls 48
