A new analysis of over half a million scientific papers reveals a rapid, near-exponential increase in the adoption of AI foundation models across various disciplines, particularly in linguistics, computer science, and engineering. The study found that open-weight models are predominantly used, and while scientists are adopting these models, they are increasingly using smaller versions compared to the larger models being developed by AI builders. This trend suggests that scientists may be missing out on the full benefits of AI, as papers utilizing larger models tend to be published in higher-impact journals and receive more citations. AI
IMPACT Suggests that the gap between AI development and scientific adoption of models may be limiting scientific progress and impact.
RANK_REASON Analysis of scientific papers on arXiv regarding AI foundation model adoption. [lever_c_demoted from research: ic=1 ai=1.0]
- Alexander Fogelson
- arXiv
- computer science
- engineering
- foundation model
- Hugging Face
- Language Models
- linguistics
- open-weight models
- Vision Models
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
