Artificial intelligence dominates headlines, attracts billions in investment and has convinced much of the world that it thinks. China expert, analyst and cultural critic Frans Vandenbosch argues that it does not, and offers the evidence to prove it.
Drawing on years of direct, hands-on experience with both American and Chinese AI platforms, Frans presents a sober and at times startling assessment of where these systems truly stand in 2026. He examines the fundamental philosophical divide between American AI, built around conversational general intelligence, and the Chinese approach, which treats AI as a practical engineering tool for industry and governance. He reviews the leading platforms on both sides, from OpenAI and Claude to Deepseek, Qwen and Doubao, assessing their real strengths and their frequently overlooked weaknesses.
At the heart of the lecture is a disturbing finding from a landmark Tencent study: when tested on tasks requiring genuine in-context learning rather than mere information retrieval, ten leading AI models achieved an average success rate of just 17.2 per cent. Bigger models, it turns out, do not learn better. They simply make more errors with greater confidence.
Frans also exposes the entrenched geopolitical bias embedded in all current AI systems, the illusion of openness surrounding proprietary platforms, and the specific conditions under which AI collapses into what the industry calls hallucination, generating fluent, authoritative and entirely fabricated nonsense.
This lecture will leave you better equipped to use AI wisely, deeply sceptical of those who claim it is on the verge of human-level reasoning, and rather less impressed by the hype than you were before you arrived.
https://youtu.be/7oKpr4o1-ow
