DeepSeek founder Liang Wenfeng's doctrine #2：
- AI currently lacks continuous learning， not taste or intuition.
- AI cannot replace human workers yet because it requires full context every time； true next-gen models must learn continuously.
- The fastest path to AGI is building models that help us develop future models， rather than prioritizing general public utility.
- Continuous learning remains unsolved because "learning" comprises many distinct processes.
- DeepSeek's AGI roadmap progresses from Chain of Thought （last year） to agents （this year）， with continuous learning as the next hurdle.
- Continuous learning will trigger a gradual singularity-where AI accelerates its own research-followed by embodied intelligence.
- Intelligence will eventually become embodied because what people truly need is physical labor， not just software.
We summarized the key remarks from DeepSeek founder Liang Wenfeng during their investor meeting. The core topics include open-source strategy, cost reduction, a...