ChatGPT Rival, Grok Goes Open Source
But the data that is open-sourced does not include Grok's training data or real-time data connections.
Elon Musk's AI company, xAI, has followed through on its plan to open-source its AI chatbot Grok to developers, as announced by Musk earlier in March.
Developers can now access Grok's base model architecture, Grok-1, a 314 billion parameter Mixture-of-Experts model, from a GitHub checkpoint last October. However, the data that is open-sourced does not include Grok's training data or real-time data connections.
This is a break away from the standard practices of industry rivals such as OpenAI, Meta, and Google, with Grok's open-source release not only permitting commercial use but also providing an opportunity for developers to delve into its intricacies and contribute to its evolution.
Now available on GitHub under the Apache 2.0 license, this move reflects Musk's push for transparency and collaboration in AI development. The Billionaire has been openly critical of companies like OpenAI for keeping their AI proprietary, despite the initial plan to function as an open-source company.
While some companies release open-source or limited open-source models to gather feedback from researchers, only a select few offer fully open-source AI foundation models like Mistral and Falcon. In contrast, the most prevalent models in the industry are either closed-source or come with limited open licenses.
Take Meta's Llama 2, for instance. While the company offers its research for free, it imposes fees on customers with 700 million daily users and restricts developers from iterating on top of the Llama 2 model.
