Science Communication

Artificial Intelligence in Scientific Publishing

A key problem facing science publishing is the firehose of data. For example, reading all monthly submissions to arXiv would take 8 years, and all papers on COVID-19 would take 150 years. To address this problem and increase the accessibility of research, we launched a platform to experiment with AI tools for data processing, including:

• AI summarization of articles
• AI-guided peer review (matching of papers and referees)
• Automatic detection of low-quality or fraudulent content
• Decomposition of papers into facts
• Similarity and contradiction indices
• Ontology (knowledge graphs) of papers enriched with AI analysis of data

We deploy these AI tools at ScienceCast.org, which is run by the company ScienceCast in Baltimore, MD.

ScienceCast Pilots

• AI writes summaries of preprints in bioRxiv trials
• Collaboration with ElevenLabs & arXiv Generates Digestible Videos for Open Access Research
• ArXivLabs collaboration links researchers’ videos to their arXiv papers

Special Issues

We ran a few special issues at Annals of Physics dedicated to the life and work of several seminal scientists, including:

• Special issue on Philip W. Anderson
• Eliashberg theory at 60: Strong-coupling superconductivity
• Special issue in memory of Prof. Konstantin B. Efetov