x.com/deredleritt3r/status/2080317020831088838
1 correction found
He says that video models and world models have no relation to the intelligence ceiling, so DeepSeek isn't interested in them.
The transcript supports a strong dismissal of video generation, but not this same categorical claim about world models. On world models, Liang says they are 'not the most important thing at this stage,' which is narrower than saying they have 'no relation' to intelligence.
Full reasoning
The post collapses two different statements from the transcript into one stronger claim.
According to the leaked transcript, Liang says:
- about video generation: it is not on DeepSeek's AGI main line, and "it has nothing to do with" the intelligence roadmap;
- about world models: they are "not the most important thing at this stage", with AI training and continual learning taking priority.
That is not the same as saying both video models and world models "have no relation to the intelligence ceiling."
So the post overstates Liang's position on world models. His comment there is qualified and time-bound ("at this stage"), not an absolute dismissal.
2 sources
- 梁文锋四小时投资人会议实录：4小时、118个回答，谁将吃到AI红利？
25.我们只做AGI的主线。AI领域很广泛，有很多东西我们觉得它不在这个主线上面，比如说3D、视频生成，我觉得可能跟智能的主线没有太大的关系，我们不会去做。26....它跟智能的路线图是没有什么关系的。28.从我们的判断，世界模型和智能还不是现在这个阶段最重要的事情。最重要的才是AI训练，以及AI训练之后怎么解决持续学习。
- Full Transcript of Liang Wengfeng's Investor Exchange: Vision, Open Source, and AGI Roadmap
'...such as 3D and video generation. I think they may not have much to do with the main line of intelligence, so we will not do them.' Later: 'From our judgment, world models and intelligence are not yet the most important things at this stage.'