What’s the news: Elon Musk’s artificial intelligence startup xAI has open-sourced some parts of its large language model (LLM) Grok-1. “We are releasing the base model weights and network architecture of Grok-1,” the company announced in a blog post.
What do these terms mean? Base or model architecture refers to “the code that specifies the structure and design of an AI model… the types of inputs and outputs to the model, how input data are processed, and how learning happens in the model.” Model weights are “the variables or numerical values used to specify how the input is transformed into the output.”
What does this mean for developers and researchers? While it has been pointed out that the model is not completely open-source, open-sourcing the base architecture and weights of the AI model will basically allow users–like developers, researchers, and business owners–to reproduce, build on it, and use it as per their requirements.
Also Read: Views: Open Source AI—A Nebulous Concept Bearing a Heavy Weight
As specified in the blog post, the model released by xAI is “not fine-tuned for any particular task” and is only “the raw base model checkpoint from the Grok-1 pre-training phase.” Additionally, the company released both the weights and the architecture under the Apache 2.0 license, meaning that while it can be used commercially, “it cannot be trademarked and there is no liability or warranty that users receive with it,” VentureBeat explained.
In the recent past, several other companies have come out with their ‘open-source’ AI models with various levels of restrictions, bringing the practice of the very usage of the term ‘open-source’ (is it just a marketing tactic?) into question. Meta’s Llama especially faced the heat for this. Meanwhile, Google came out with its Gemma and then there’s Mistral AI that some consider to be Europe’s version of OpenAI.
Also Read: Navigating the Complexities of Open Source AI: Insights from Carnegie India Summit
Right from the start, Musk positioned Grok-1 as a direct competitor of OpenAI’s ChatGPT 3.5 stating that Grok-1 LLM “surpasses” ChatGPT3.5 and ChatGPT4 in its capabilities. The open-sourcing of the model also comes against the backdrop of a heated legal battle between Musk and OpenAI (of which Musk was a co-founder) in which the former has alleged the latter of betraying the “founding agreement” signed between Musk and OpenAI’s Sam Altman and Greg Brockman. Musk has argued that by licensing GPT-4, OpenAI CEO Sam Altman has departed from OpenAI’s “original mission and historical practice” of open-sourcing its technology. You can read more about the lawsuit here.
Also Read:
- Elon Musk’s xAI launches AI chatbot Grok that has access to real-time info from X
- Elon Musk Sues OpenAI For Violating Founding Agreement of Developing Artificial General Intelligence For Public Benefit
STAY ON TOP OF TECH NEWS: Our daily newsletter with the top story of the day from MediaNama, delivered to your inbox before 9 AM. Click here to sign up today!