Large-scale generative simulation artificial intelligence: The next hotspot

Practical considerations and the roadmap of LS-GenAI

Nowadays, big data, deep-learning models, optimization methods, and computational power are essential elements in promoting the development of artificial intelligence. Recent advances have brought a new focus on generative artificial intelligence (GenAI), which paves promising paths to exploring the creation of texts, images, videos, or other content rather than simply performing discriminative learning tasks.

GenAI’s emergence, e.g., the large model, has changed the landscape of deep-learning research and has influenced individuals' work and life, holding tremendous potential to reshape robotics research, national governance, and life sciences. Consequently, a pressing question arises: “How will GenAI inspire technological revolution?” In answer to this question, this commentary highlights the fascinating functions of GenAI, stresses the importance of experimental design, discusses critical inductive biases as the geometric prior, identifies multi-views in the evaluation, and nominates large-scale generative simulation AI (LS-GenAI) as the next hotspot for GenAI to connect.