Physical Address

304 North Cardinal St.
Dorchester Center, MA 02124

Meta’s LlamaCon was all about undercutting OpenAI


On Tuesday, Meta, the EU Developer Conference, Llamacon, his Menlo Park, California headquarters in the first place. The company has announced Starting a Meta AI chatbox application to consumer-faceIt will also compete with ChatrEpt, but also a Developer-Facial API To access the llama models in the cloud.

Both releases aims to expand the adoption of open LLA AI models, but this goal may be secondary to the real motif of the meter: to beat Openai. Meta’s AI fame, in large strokes, such as Openai, which went out of the door behind the services, turns on the evolving open AI ecosystem.

The AI ​​chatbot application of methane is almost a privilege Openai’s rumor social network. There is a social dish that users can share AI conversations and a user has a social dish that has submitted individual answers based on the meta app activity.

As for the Llama API, Openai’s API is a problem. The Llama API is designed to make the developers simpler to build applications in the cloud, using only one code line. The third party to operate the Models of LLA allows the need to trust cloud providers and allows Meta AI to offer a more full range for developers.

Meta, as many AI companies, accept that Openai is a top opponent. In court the trial, the meta revealed the execution of the company previously obsessed Once upon a time, Openai’s GPT-4 beating with a modern model. The placement of ownership AI model providers such as Openai is a long time to the AI ​​strategy of meters. One July 2024 letterMeta CEO Mark Zuckerberg wrote that he wrote “access to AI models” wanted to compare meta with companies such as Openai [Meta’s] work model. “

Several researchers who talked to the pioneers in front of Llamaco were hoping for meta Leave a competitive AI thinking model like Openai O3-Mini. The company did not end to do it. However, for meta, it is absolutely not going to win the AI ​​race.

During the stage during Llamacon, Dateberg, Deepsek and Alibaba’nın qwen, including Ghodsi, Deepseek and Alibaba’nın qwen, said he saw any AI lab against open model providers.

“A part of the value around the Open Source is that you can mix and match. Thus, if Gwen is better, you have the ability to take the best parts of the intelligence from different models,” Zuckerberg. “This is a part of what I think of how the open source is mainly in the quality of a closed source [models] … [I]t feels like a non-stop type of force. “

In addition to stopping Openai’s growth, Meta is trying to push open models To ensure regulatory caravates. AU Ai ACT provides special benefits to companies distributing AI systems “free and open source”. Meta often claims the “open source” of the llama models Although disagreeing to not meet the necessary criteria.

Regardless of the reason, opening the EU’s open model ecosystem to open an AI launcher and restricting Openai’s growth, sometimes the generations of advanced models themselves are generally.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *