Abstract
Multiple firms spanning expertise from material science and computational chemistry to pharmaceuticals have started investing into Artificial Intelligence (AI) and Large Language Models (LLMs) to streamline and expedite their Research and Discovery (R&D) process; however, the effects of this newfound integration of AI into R&D remain unclear as it is yet to be understood whether it is changing how the process is conducted versus simply accelerating the process. This study utilizes qualitative data collected from seven semi-structured interviews with interviewees with expertise spanning from corporate research and startups to innovation economics in order to analyze the central question: does the integration of AI in R&D deepen research in already explored areas by accelerating it or does it widen the directions of research and introduce novel directions? Across all interviews, the most common limitation noted in AI’s integration in R&D was human validation. Regarding the central question, the results skewed towards the deepening trend. Many interviewees mentioned that the current usage of AI causes its integration to only accelerate and improve the discovery within research directions already explored; however, interviewees explained that this was a trend based on the methodology of R&D used by most firms rather than a technological barrier of the tool itself. These results suggest that firms and policymakers should thoroughly evaluate the productivity increase of AI rather than its potential and focus on solving the several bottlenecks still not tackled by AI: validation, human judgement, and physical experimentation.