On July 4, at the 2024 World Artificial Intelligence Conference and High level Conference on Global Governance of Artificial Intelligence Industry Development Main Forum, Robin Lee, the founder, chairman and CEO of Baidu, once again talked about the battle between the open source and closed source routes of the big model.
He stated that open source big models have value in specific scenarios such as academic research and teaching, but they are not suitable for most application scenarios. "When you are in a fiercely competitive environment and need to make your business more efficient and cost-effective than your peers, a commercial closed source model is the most effective."
Robin Lee said that since this year, open source and closed source big models have been a controversial topic, but many people have confused the concepts of model open source and code open source. "Open source models can only get a bunch of parameters, and they also need to do SFT and secure alignment. Even if you get the corresponding source code, you don't know how much proportion or what proportion of data was used to train these parameters, which cannot achieve 'everyone gathers firewood and flames high', and cannot make you iterate on the shoulders of giants."
Robin Lee said frankly that under the same parameter scale, the open source model is not as powerful as the closed source model
Robin Lee said that the open source model is valuable in some academic research and teaching fields and can be used to study the working mechanism of the big model and form a theory. However, open-source models are not suitable for most application scenarios. In a fiercely competitive business environment, if you want to achieve higher business efficiency and lower costs than your peers, a commercial closed source model is the most effective. He took Baidu's practice in novel creation as an example, and after transitioning from open-source models to lightweight models and then to Wenxin Big Model 4.0, the availability and quality of novel generation have greatly improved.
Robin Lee believes that the focus of the big model is still "application". "Without application, only the basic model, whether open source or closed source, is worthless."