Search for: All records

Large-language models (LLMs) such as GPT-4 caught the interest of many scientists. Recent studies suggested that these models could be useful in chemistry and materials science. To explore these possibilities, we organized a hackathon. This article chronicles the projects built as part of this hackathon. Participants employed LLMs for various applications, including predicting properties of molecules and materials, designing novel interfaces for tools, extracting knowledge from unstructured data, and developing new educational applications. The diverse topics and the fact that working prototypes could be generated in less than two days highlight that LLMs will profoundly impact the future of our fields. The rich collection of ideas and projects also indicates that the applications of LLMs are not limited to materials science and chemistry but offer potential benefits to a wide range of scientific disciplines. 

Abstract Numerical models of ocean biogeochemistry are relied upon to make projections about the impact of climate change on marine resources and test hypotheses regarding the drivers of past changes in climate and ecosystems. In large areas of the ocean, iron availability regulates the functioning of marine ecosystems and hence the ocean carbon cycle. Accordingly, our ability to quantify the drivers and impacts of fluctuations in ocean ecosystems and carbon cycling in space and time relies on first achieving an appropriate representation of the modern marine iron cycle in models. When the iron distributions from 13 global ocean biogeochemistry models are compared against the latest oceanic sections from the GEOTRACES program, we find that all models struggle to reproduce many aspects of the observed spatial patterns. Models that reflect the emerging evidence for multiple iron sources or subtleties of its internal cycling perform much better in capturing observed features than their simpler contemporaries, particularly in the ocean interior. We show that the substantial uncertainty in the input fluxes of iron results in a very wide range of residence times across models, which has implications for the response of ecosystems and global carbon cycling to perturbations. Given this large uncertainty, iron fertilization experiments based on any single current generation model should be interpreted with caution. Improvements to how such models represent iron scavenging and also biological cycling are needed to raise confidence in their projections of global biogeochemical change in the ocean.