Tyler R. Josephson 藍泰來 (he/him)
LightNews LightNews
banner
atomslab.bsky.social
Tyler R. Josephson 藍泰來 (he/him)
@atomslab.bsky.social
210 followers 300 following 23 posts
AI & Theory-Oriented Molecular Science. Asst Prof at UMBC, molecular simulations for the environment + automated reasoning for chemical science and engineering. RT≠PV/n

https://atomslab.github.io
Posts Replies Media Videos
atomslab.bsky.social
Haha, me too!
August 2, 2025 at 2:53 PM
atomslab.bsky.social
This is one of my favorite slides from Lean for Scientists and Engineers. Traditionally, people imagine new scientific theories, derive them by hand, and implement them in software as best as they can, then compare to experiment. But this is hard, and people can make mistakes along the way!
Flowchart illustrating relationship among pure math, scientific models, scientific software, and reality. In the realm of pure math, syntax separates logically valid statements from all possible combinations of symbolic. Logically valid statements bridge into symbolic scientific models, which also need to bridge to computable models. Computable models are implemented as code in scientific software, which finally bridges to reality. A final bridge does connect symbolic models directly to reality, but the main path is via computable models and code. The previous chart, with three boxes overlaid. Traditionally, the validity of the mathematics and the scientific theory are established by hand. Humans read the theory and write the code as best as they can. Then use various automated and manual means to compare to experiment. The three overlaid boxes have been replaced with just two: Can we represent all of this in Lean, and validate the construction of the math, scientific models, and software, in one system? Then use various automated and manual means to compare to experiment.
February 25, 2025 at 4:17 PM