Alibaba seeks to compete with its AI models, with those already existing in the market
Alibaba released several days ago, has released two major language models (LLM) open source in an attempt to deal with metaplatforms.
artificial intelligence models, called Qwen-7B and Qwen-7B-Chat, which each model has seven billion parameters, which are often used to measure its strength. One of it's AI models from Alibaba, seeks to position itself as a conversational variation to Meta's recent release of "Llama 2".
The company said in a press release that the "code, model weights, and documentation for the two models will be freely accessible to academics, researchers, and commercial institutions worldwide.«. This is the first time that a major Chinese technology company has made its LLMs open access.
He added that the launch of these open source models aims that these great LLM models can be part of the effort for helping democratize AI technology and support LLM start-ups. In addition, it was also mentioned that free access to the code, documentation, and internals of the models will be provided to academics, researchers, and commercial institutions around the world.
It must be remembered that LLM AI models can process massive amounts of data and generate all kinds of content from text, images, audio and video, in addition to the information serving as the backbone of AI chatbots such as ChatGPT.
About their LLM models, Alibaba He mentioned that the two are based on Alibaba's Tongyi Qianwen LLM, which was announced in April. Tongyi Qianwen has multiple versions with different numbers of parameters, and Qwen-7B and Qwen-7B-Chat are two small-sized versions that are designed to help small and medium-sized businesses get started with AI.
"Our goal is to promote inclusive technologies and enable more developers and small and medium-sized businesses to reap the benefits of generative AI," said Alibaba Cloud Intelligence CTO Zhou Jingren.
While major tech companies are "open sourcing" their AI models, openness is up for debate. Like Alibaba, Meta admits that there are parts of LLaMA 2 that cannot be accessed. Similarly, OpenAI limits what you share from your GPT-4 AI. A recent study shows that, in fact, OpenAI's ChatGPT and Meta's LLaMA are not very open at all.
Alibaba Cloud said that companies with fewer than 100 million monthly active users are free to implement the models; while companies with more than 100 million monthly active users will need to apply for a license from Alibaba before using both models. Similarly, Meta's Llama 2 requires a license for users of more than 700 million users.
Alibaba said Tongyi Qianwen had versions with more parameters that were not released for free, but did not say how many parameters its largest model contained. The largest version of Meta's Llama 2 has 70 billion parameters.
“Today, Alibaba Cloud is the first to launch a training and implementation solution for the entire Llama 2 series in China,” Alibaba posted on its WeChat channel. "All developers can create custom large models on Alibaba Cloud."
China tries to catch up with the United States in the field of AI, and Beijing encourages Chinese companies to rapidly develop domestic and “controllable” AI models capable of rivaling those developed by American companies.
In addition to Alibaba, many of the tech giants, such as Huawei, have been aggressively developing their own AI models in recent months. Fountain
As already mentioned, for those interested, the models based on Alibaba's Tongyi Qianwen (Qwen-7B, and the Qwen-7B-Chat version) for conversations are available to anyone with a free license and to companies with more than 100 million monthly users, one must be requested.
Finally if you are interested in knowing more about it, you can check the details in the following link