AI Three Questions ③ Model Inquiry | Facing Model Challenges with Love to Shape the Future of AI — WAIC 2025 Large Model Forum Uses Questions to Drive Innovation}

WAIC 2025

World Artificial Intelligence Conference

Forum: July 26-28, 2025

Exhibition: July 26-29, 2025

Locations: Expo Center, Expo Pavilion, Xuhui West Bund, etc.

To delve into AI’s core topics, WAIC boldly proposes the “Three Questions of AI”—focusing on the frontiers of mathematics, science, and models. The mathematics question explores axioms and formulas to build cognitive frameworks; the science question roots in empirical evidence to uncover natural laws; the model question integrates both, transforming abstract concepts into practical tools.

These three questions are interconnected: mathematics quantifies science, science gives mathematical meaning, and models realize intelligence. Their synergy drives innovation across diverse fields.

The 2025 WAIC will be held in Shanghai from July 26-28. As a globally influential AI industry event, WAIC promotes technological breakthroughs, industry integration, and global governance dialogues. On July 27 morning, the “Boundless Love & Shaping the Future” Large Model Forum, hosted by SenseTime, will open at the Shanghai Expo Center. Centered on core questions, the forum will explore deep issues of models, fostering global wisdom to shape AI’s future.

Highlight 1: Cross-border, Cross-architecture Dialogue to Build a Thinking High Ground with “Model Questions”

This key segment aims to create a global platform for top research and technical exchange, focusing on “solving the fundamental problems of models.” Experts from leading AI companies and top universities will discuss whether the limitations of generalization stem from architecture and learning paradigms, exploring breakthroughs. Diverse perspectives will collide and merge, promoting frontier achievements and providing multiple viewpoints to overcome current technical bottlenecks, making “model questions” the starting point for innovation.

Highlight 2: Architectural Innovation and Industry Application to Drive Paradigm Shifts with “Model Questions”

The event will focus on integrating Transformer and non-Transformer architectures, promoting multi-paradigm evolution. Topics include addressing the semantic gap in multimodal intelligence—solving the semantic mismatch between text and images—and optimizing performance-cost curves by reducing training energy and computational demands through lightweight and differentiated architectures. Academic discussions will target “performance limits during inference,” analyzing how model size and complexity affect performance under reinforcement learning paradigms. Industry presentations will showcase engineering solutions, fostering a synergy between research and application, and making “model questions” a bridge for sustainable AI development.

Highlight 3: Building Global Consensus to Guide Development with “Model Questions”

This event gathers industry leaders, scholars, and researchers to discuss future trends and paths for large models. Led by SenseTime’s co-founder, top Chinese companies and academics will analyze the core obstacles to high-level intelligence, such as complex reasoning and autonomous decision-making, exploring technical breakthroughs to develop “Chinese wisdom” solutions. International experts will also discuss innovations in architecture and computing power, focusing on bridging semantic gaps and breaking performance boundaries. Deepening global collaboration, “model questions” will become a key theme guiding industry progress, shifting AI from competition to joint problem-solving, and providing clear questions and new approaches for global AI advancement.

      ^{Scan the QR code to register for the forum live broadcast}