Speech by Academician Sun Ninghui on the Development, Challenges, and Future of Artificial Intelligence and Intelligent Computing in China

01 Computing Technology Development Overview – Human computation began with the abacus and progressed through mechanical, electronic, and network eras to the current fourth stage of intelligent computing. Key milestones include Babbage’s analytical engine, Ada Lovelace’s algorithms, Boolean algebra, Turing machines, and the von Neumann architecture, culminating in modern platforms such as high‑performance computing, servers, PCs, smartphones, and embedded systems.

02 Intelligent Computing Development Overview – Intelligent computing has evolved through four phases: general‑purpose computers, expert‑system reasoning, deep‑learning systems, and today’s large‑model era. The emergence of ChatGPT and other large models (Gemini, LLaMA, SAM, SORA) has shifted AI from discriminative to generative, driving massive data, algorithm, and compute demands.

03 AI Safety Risks – Rapid AI progress brings security concerns: deep‑fake avatars, fabricated videos, synthetic news, voice‑cloning fraud, illicit imagery, and model hallucinations. Trustworthiness issues include factual errors, political bias, inducible harmful outputs, and data privacy risks. Legislative responses span China’s AI ethics standards, machine‑learning safety guidelines, and international regulations such as the EU GDPR, US AI Bill of Rights, and the EU AI Act.

04 China’s Intelligent Computing Development Challenges – China faces four major hurdles: (1) lagging behind the US in core AI talent, algorithms, and foundational models; (2) restricted access to high‑end chips (A100, H100, B200) due to export bans; (3) a weak domestic AI ecosystem with limited CUDA‑compatible tools and talent; (4) high cost and barriers for AI adoption across industries.

05 China’s Path Choices for Intelligent Computing – Three strategic routes are proposed: (A) chase the US‑dominated “A‑system” by aligning with CUDA and existing ecosystems; (B) build a closed, specialized “B‑system” for defense, meteorology, and other verticals using domestic chips; (C) develop an open, global “C‑system” based on open‑source hardware and software (e.g., RISC‑V, open AI models) to foster shared infrastructure, data, and models. Additionally, the nation must decide whether to prioritize algorithmic advances or invest in new data‑center and compute infrastructure, and balance AI’s role in virtual versus real economies to ensure sustainable, high‑quality growth.

The speaker, a member of the Chinese Academy of Engineering and a senior researcher at the Institute of Computing Technology, Chinese Academy of Sciences, emphasizes the need for affordable, trustworthy AI that supports both domestic and international development goals.