Kuai Technology reported on April 12 that in a recent internal speech, Li Yanhong responded for the first time to why the Wenxin large model is not open source.
Li Yanhong stated that there was a very intense discussion when Wenxin was first released a year ago, but they ultimately decided not to open source it because the judgment at the time was that there would definitely be more than one open-source option in the market.
“In this situation, having one more open-source option from Baidu doesn’t make much difference, and having one less doesn’t make much difference either,” Li Yanhong said.
Li Yanhong also believes that closed source has a real business model, which can make money, and making money can gather computing power and talent.
At the same time, closed source has cost advantages. Given the same capabilities, the inference cost of a closed-source model is definitely lower, and the response speed is definitely faster.
Li Yanhong mentioned that, “Whether in China or the US, the current strongest foundational models are closed source, and the models made by reducing the dimensionality of these foundational models are also better, which makes closed source more advantageous in terms of cost and efficiency.”
Will Wenxin Yiyan become Wenxin’s last words?
Baidu not doing evil is intolerable by heaven.
Personally, I don’t recommend disparaging others. Expressing opinions doesn’t really have much significance, and the audience isn’t on the forum. It might be more effective to make a short video instead.
Moreover, looking at it rationally, TiDB and Wenxin Yiyan are doing two different things.
One is a model, and the other is a vector database. They shouldn’t be compared together. If Wenxin Yiyan is to be compared in the open-source domain, it should be compared with something like ChatGLM.
When it comes to comparing vector databases, there are many more competitors.
Look at it rationally. Seek truth from facts.
Actually, whether it’s open source or not doesn’t necessarily determine a bigger future. The level of achievement depends on the comparison. Moreover, the two are not the same product and are used in completely different scenarios. Be mindful of your wording online, especially when harshly criticizing other products on one platform.
From the user’s perspective, everyone is more concerned about the user experience of the product. Under the same conditions, if the open-source version can meet the requirements, I believe no one would spend money on the commercial version, especially when the user experience of some closed-source products is uncertain or not good.
From the user’s perspective, open source indeed takes away a lot of markets that previously required purchasing enterprise services to make money. Open source is not conducive to enterprise commercialization. However, by leveraging the differences in various regions of the world, China can develop open source, quickly iterate and refine products, while the West can make money and use closed source to support open source. TiDB is the first to do this.
That’s the logic. If a company wants to go far, it will definitely be difficult to sustain without a stable source of income. Either commercialize or have traffic, so that companies will sponsor you. If operated well, relying on website advertising revenue or corporate sponsorship in the later stages is also a viable path.
|
username: tidb狂热爱好者
|
username: 
|
username: 
|
username: 
Be mindful of your wording online, especially when harshly criticizing other products on one platform.
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: