{getToc} $title={Table of Contents}
Summary
The study compares four approaches to identifying patented AI inventions: Keyword, Science, WIPO, and USPTO. The results show that the approaches overlap in only 1.37% of patents and vary in size, with the USPTO approach identifying the largest number of patents. The study evaluates the "GPTness" of AI patents using growth, generality, and complementarity metrics, finding that all methods classify AI as a GPT, but with varying degrees of GPTness.
Highlights
- Four approaches to identifying AI patents: Keyword, Science, WIPO, and USPTO.
- Approaches overlap in only 1.37% of patents.
- USPTO approach identifies the largest number of patents.
- All methods classify AI as a GPT, but with varying degrees of GPTness.
- Keyword method captures the highest level of GPTness.
- Concentration of AI inventions among top firms is a concern.
- Results have implications for AI policy and research.
Key Insights
- The study highlights the challenges of defining and measuring AI, with different approaches yielding varying results. This underscores the need for a nuanced understanding of AI and its applications.
- The findings suggest that AI is a general-purpose technology (GPT), but the degree of GPTness varies across approaches. This has implications for AI policy and research, as GPTs are often associated with significant economic and social impacts.
- The concentration of AI inventions among top firms is a concern, as it may limit the diffusion of AI technologies and exacerbate existing inequalities. Policymakers may need to consider measures to promote greater diversity and inclusivity in AI development.
- The study's results have implications for AI research, as they highlight the importance of considering multiple approaches and perspectives when studying AI. This can help to ensure that research is comprehensive and nuanced, and that findings are robust and generalizable.
- The Keyword method's high level of GPTness suggests that it may be a useful approach for identifying AI patents, particularly in contexts where precision is paramount. However, the limitations of this approach should also be considered, as it may not capture the full range of AI-related inventions.
- The study's findings highlight the need for ongoing research and analysis to better understand the complex and evolving landscape of AI. This can help to inform policy and decision-making, and ensure that the benefits of AI are realized while minimizing its risks and challenges.
- The results of the study have implications for the development of AI policies and regulations, as they highlight the need for a nuanced and multi-faceted approach to AI governance. This may involve considering multiple perspectives and approaches, and engaging with a range of stakeholders to ensure that policies are effective and responsive to the needs of different groups.
Mindmap
Citation
Hötte, K., Tarannum, T., Verendel, V., & Bennett, L. (2022). Measuring artificial intelligence: a systematic assessment and implications for governance (Version 3). arXiv. https://doi.org/10.48550/ARXIV.2204.10304