The following guide announces China’s funding priorities for explainable and generalizable AI basic research in 2025. The Chinese government is funding cutting-edge basic research on various approaches to AI, including generative large models but also brain-inspired and cognitive AI models. The guide also encourages and subsidizes novel applications of AI in physical modeling, biological and health research, and materials science.

An archived version of the Chinese source text is available online at:    https://perma.cc/EN2E-HLP8

Guide to the 2025 Annual Projects for the Major Research Program on Explainable and Generalizable Next-Generation Artificial Intelligence Methods

The Major Research Program on Explainable¹ and Generalizable Next-Generation Artificial Intelligence Methods (“the Major Research Program”) is geared towards: The major national strategic requirements of artificial intelligence (AI) development, taking the basic scientific problems of AI as the core; developing a system of new AI methods; and promoting China’s basic AI research and talent cultivation, thereby supporting China’s dominant (主导) position in the new round of international competition in science and technology (S&T).

I. Scientific Objectives

The Major Research Program is geared towards: Poor robustness, poor explainability, strong data dependence, and other basic scientific problems in AI methods as represented by deep learning; uncovering the basic principles of machine learning; and developing explainable and generalizable next-generation AI methods, so as to promote the innovation and application of AI methods in scientific fields.

II. Core Scientific Problems

The Major Research Program is aimed at the basic scientific problems of explainable and generalizable next-generation AI methods, centered around conducting research on the following three core scientific problems.

(i) Basic principles of deep learning.

Deeply uncover the dependency of deep learning on hyperparameters, understand the working principles behind deep learning, and, for deep learning methods, establish approximation theory, generalization error analysis theory, and the convergence theory of optimization algorithms.

(ii) Explainable and generalizable next-generation AI methods.

By combining rules and learning, establish new AI methods that are high-precision, explainable, and generalizable, and do not rely on large amounts of annotated data. Develop the databases and model training platforms needed for next-generation AI methods, and improve infrastructure driven by next-generation AI methods.

(iii) Applications of next-generation AI methods geared towards scientific fields.

Develop new physics models (物理模型) and algorithms, build open-source scientific databases, knowledge bases, physics model libraries, and algorithm libraries, and promote exemplary applications of new AI methods in solving complex problems in scientific fields.

To view the rest of this translation, download the pdf below.

1. Translator’s note: In the context of AI, the Chinese word 可解释性 can be translated as either “explainable” or “interpretable.” This translation opts for the former.Footnote Link