In his words
Constantly redeveloping foundational large models represents an enormous waste of social resources.
I’ve noticed that many people still primarily focus on foundational models. But I want to ask: how about real-world applications? Who has benefited from them?
Biography
Robin Li (born 1968) co-founded Baidu in 2000 after inventing RankDex, a link-analysis search algorithm that predated Google’s PageRank, and built it into China’s dominant search engine. He began repositioning Baidu as an AI company in the mid-2010s — autonomous driving, voice assistants, and the deep-learning research that would become the Ernie model family — long before it was fashionable.
Ernie Bot and the applications gospel
In March 2023 Li launched Ernie Bot, China’s first major answer to ChatGPT, which grew to hundreds of millions of users. But his signature message has been contrarian within the boom: China’s “war of a hundred models” squanders compute, and value comes from AI-native applications, not from every company training its own foundation model. At the 2024 World AI Conference in Shanghai he needled an industry obsessed with benchmarks to show him the applications and the beneficiaries; at Baidu World 2025 he argued applications must ultimately create a hundred times the value of the models beneath them.
The evangelism has had mixed returns for Baidu itself: Ernie lost the consumer chatbot race to ByteDance’s Doubao and DeepSeek’s app, and in 2025 Li reversed a years-long defense of closed models by open-sourcing Ernie — a U-turn that followed Liang Wenfeng’s open-source shockwave. Li remains one of AI’s steadiest bulls, telling audiences that hallucinations are fading and that AI will amplify rather than erase human work.
Where they stand in the war
Who backs them up
Sources & further reading
Canonical record: https://battlelines.ai/topic/robin-li








