Meta suddenly re-embraces open source AI. Zuckerberg expresses his position: Distillation is not guilty.

Kuai Technology reported on August 10 that Meta company was once the leader in the global open source AI model, and then chose closed source. Now they have changed again and embraced open source again. Zuckerberg also set the tone for distillation, emphasizing that distillation is not guilty. Recently, Meta has released a number of large models. The performance of the latest Muse series has been significantly improved. Even if users choose to hand over data to Meta, the price can be reduced by up to 95%, which is cheaper than DS V4 Flash. American large models are also going to be cost-effective. Tonight, Meta has fulfilled its promise and officially opened the Muse series model weights. The first one is Muse Glimmer, a 30B parameter Agent model, distilled from Muse Spark, which can be run locally on a Mac or a PC with a single graphics card. Zuckerberg also officially confirmed that the weight of the main large model Muse Spark 1.2 will also be opened in the next few weeks. Although there was a gap of four months between the open model weight and the promised time, Meta finally did not break its promise. Zuckerberg also issued an open letter while open source, and began to strongly support open source large models, especially to express his position on the controversial distillation issue. Zuckerberg said that models learning from other models is part of the open source ecosystem. If the United States wants to continue to lead open source AI, it must relax restrictions on distillation and training data. He opposed the statement that distillation is harmful, emphasizing that distillation is innocent, and believes that what can be observed should be allowed to be learned. If the United States restrains itself first, it will only eventually make other countries' open source models catch up faster. His views on distillation are different from those of OpenAI and Anthropic. In order to maintain their status, the latter two strongly opposed other companies' distillation of their own big models, and repeatedly lobbied the U.S. Congress to classify distillation as a crime. Their attitude towards distillation is extremely hostile and crazy, without mentioning that their big models themselves are distilled from the accumulation of all human knowledge and the Internet, including huge Chinese materials. Of course, although Zuckerberg’s attitude towards distillation is different from this