How Large Models Are Redefining AI and Shaping the Next Industrial Revolution

Robin Li, founder and CEO of Baidu, delivered a keynote titled “Large Models Change the World” at the 2023 Zhi Guan Cun Forum, arguing that we are at the beginning of a new AI era where large models become the core of innovation and will soon transform the world.

Large Models Redefine Human‑Computer Interaction

Li emphasizes a shift from command‑line and graphical interfaces to natural‑language interaction, creating a "prompt engineering" revolution. He predicts that within ten years, 50% of jobs will involve prompt engineering, and that the ability to formulate effective prompts will outweigh coding skills in determining compensation.

Impact on Marketing and Customer Service

According to Li, the entity that masters AI‑driven communication will dominate customers. Large models enable each of billions of users to have a 24/7, knowledgeable AI assistant, fundamentally changing how businesses engage with clients.

AI‑Native Applications

Li cites examples such as DoNotPay (legal assistance), Jasper (content generation), and Speak (language learning) to illustrate AI‑native apps that act as personal tutors or assistants. Baidu is rebuilding all its products with an AI‑native mindset, aiming to redesign rather than merely integrate AI.

Evolving Technology Stack

The traditional three‑layer IT stack (chip, OS, application) is expanding to a four‑layer model in the AI era: chip (GPU‑focused, e.g., Baidu’s Kunlun chips), framework (deep‑learning frameworks like PaddlePaddle, PyTorch, TensorFlow), model (large language models such as ERNIE Bot), and application (AI‑native services). Baidu claims end‑to‑end optimization across these layers, improving efficiency dramatically.

Baidu’s Progress and Vision

Since 2019, Baidu has released successive versions of its large model (ERNIE 1.0 → 2.0 → 3.0) and launched ERNIE Bot in March 2023, achieving a ten‑fold increase in QPS and a ten‑percent reduction in inference cost. Li attributes Baidu’s success to long‑term technical investment, a unique four‑layer architecture, and autonomous control over data, frameworks, and models.

Future Outlook and Governance

Li warns that while AI will boost efficiency, it also raises concerns about control and societal impact. He advocates international collaboration to establish governance rules, emphasizing that participation in rule‑making is essential for global competitiveness.

Overall, the speech presents a comprehensive analysis of how large AI models are reshaping interaction paradigms, business models, technology infrastructure, and economic development, positioning them as the driving force behind the next industrial revolution.