We have already talked about the language model LLaMa signed Meta, also called Large Language Model Meta AI and designed to be much lighter than OpenAI solutions. Already available on GitHub, it can be used on PCs equipped with a sufficiently performing CPU. As an open source AI model it has great promise and could deliver hard time to the alternatives proposed by OpenAI and Google. But will it really be like this? Let’s analyze the difference between Meta’s approach and that of the two rivals most intrigued by artificial intelligence.
Meta aims for an open source language model
The analysis conducted by ShotgunProxy on Reddit clearly explains the effectiveness of this approach by Mark Zuckerberg and associates. The offer of one commercial license for an open source LLM is great, as it allows developers and companies to freely adopt and profit from the AI model. By doing so, LLaMa can truly become the most popular open source solution in use.
Today it is preferred by research experts precisely because of its open nature; in reverse, GPT and the underlying Google Bard model are closed-source.
For this reason Meta might pose a considerable pressure on the two named rivalspreferring direct relationship with developers and companies rather than earning money by offering limited access to a closed-source solution.
On the one hand, OpenAI is seeing the effectiveness of this idea and is planning the release of its own open source model, although it will probably be a less performing alternative to GPT-4. On the other hand, Google seems more willing to go its own closed-source route.
Meanwhile, the US government is surprisingly moving against Meta as, in their opinion, the open source approach is “irresponsible and dangerous”. Which front will prevail and gain the support of developers and experts?