However, the use of specialized AI comes with challenges, including data security, the need for sufficient and relevant data, and the optimization of AI models to manage multiple scientific principles. Despite these challenges, Sanghavi believes that specialized AI has the potential to transform areas ranging from environmental sustainability to product innovation to human health.
Key takeaways:
- Generative AI has made significant advancements, but another approach to AI, which applies a different approach to scientific discovery, is emerging. This approach uses small, carefully curated use-case-specific AI models to speed up scientific discovery.
- Traditional scientific discovery is often cost-prohibitive, time-consuming, and limited in scope. AI in science offers a more efficient approach, potentially saving months or even years of research and development time.
- Science-based AI models, or small science models (SSMs), are trained on smaller, science-based, mostly private data models. These models can be used to investigate complex problems faster, even with smaller data sets.
- While science-based AI presents many opportunities, it also poses challenges, including data security, the need for sufficient data to fine-tune discovery capabilities, and the need for optimization to manage multiple scientific principles across different disciplines.