Meta repent? The Llama 2 model is not completely free or charged to large cloud service providers

Source: "Science and Technology Innovation Board Daily"

Editor: Song Ziqiao

Llama2 is open, but in addition to making the technology open source, Meta CEO Mark Zuckerberg said on Wednesday's earnings call that Meta includes a clause that major cloud computing companies won't get free access license, they must enter into a commercial agreement with Meta. **

"Companies like Microsoft, Amazon, or Google basically get their revenue from reselling services, and I think we should get some portion of the revenue for ... I think in the short term it's It’s not going to generate a lot of revenue**,” Zuckerberg said.

Llama 2 is the latest large model of Meta, with three parameter scales of 7 billion, 13 billion and 70 billion. Among them, the capacity of the 70 billion parameter model is close to that of GPT-3.5. Compared with Llama1 released in February this year, the token used for Llama2 training has doubled to 2 trillion. At the same time, Llama2 has also doubled the most important context length limit for using large models.

The significance of this model is not only open source, as an alternative to closed source models.

Previously Meta announced that Llama2 can be used for free for research and commercial purposes. On July 18, Meta released Llama2 and announced that it will cooperate with Microsoft and Amazon to provide global developers with cloud services based on the Llama2 model for commercial use without charging access or usage fees.

Jim Fan, a senior AI scientist at Nvidia, once spoke highly of Meta's "free" decision: the training cost of Llama2 may exceed $20 million. Meta does an incredible service to the community by releasing models with a commercially friendly license. Previously AI researchers at big companies were wary of Llama1 due to licensing issues, but now I think many of them will jump in and contribute.

Xia Qingying, an analyst at Wanlian Securities, said that Llama2, as a free and commercially available open source large model, will accelerate the popularization of large model technology. Enterprises and individuals can develop their own exclusive large models at a lower cost based on Llama2. For application developers, Llama2 greatly reduces the development threshold of AI applications, which is expected to promote the flourishing of downstream applications.

Now it seems that **Meta does not intend to do charity. Although it will not charge a large number of small developers, it will set limits for large cloud service providers. **

In fact, Meta has already determined its strategy - "be kind" to small factories and "strict" to big factories. In the previous agreement, although Llama2 allowed commercial use, it also added an additional commercial clause:

You must apply to Meta if, on the date of the release of the Llama2 version, the number of monthly active users of products or services offered by licensee or licensee's affiliates exceeds 700 million in the previous calendar month** License, Meta may, in its sole discretion, grant you that right, and you shall have no right to exercise any rights under this Agreement unless or until Meta expressly grants you such rights.

Many people think that there is nothing wrong with Meta's approach. Some netizens said, "I am completely on the side of Meta. They have done a lot of work for the open source of the large model, which is very reasonable-almost free for everyone, except for those giant companies."

Some netizens agree that Meta charges large cloud service providers

View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
0/400
No comments
Trade Crypto Anywhere Anytime
qrCode
Scan to download Gate App
Community
English
  • 简体中文
  • English
  • Tiếng Việt
  • 繁體中文
  • Español
  • Русский
  • Français (Afrique)
  • Português (Portugal)
  • Bahasa Indonesia
  • 日本語
  • بالعربية
  • Українська
  • Português (Brasil)