Journal of Artificial Intelligence, Virtual Reality, and Human-Centered Computing
Beyond Data: How Simulation Can Empower LLMs to Revolutionize Science
Abstract
M. Garbey
We hypothesize that large language models (LLMs) can effectively generate ‘thought experiments’ that can be evaluated using virtual simulators. This approach has the potential to automate knowledge generation, accelerate scientific discovery, and provide a systematic method for training subsequent LLMs with verifiable and reproducible datasets. To illustrate this concept, we will use the clinical workflow optimization challenge as a case study.
