Following ChatGPT's success, OpenAI may be releasing an open-source AI model soon

  • 2 min read
  • May 18, 2023
Capitalizing on the popularity of the GPT-powered ChatGPT, OpenAI may be looking to release an AI model that is open-source.

Text reading ChatGPT over an image of a computer screen displaying lines of code

Generative AI is all the rage these days, spearheaded by publicly available applications such as ChatGPT, Bing Chat, Google Bard, Midjourney, Stable Diffusion, and more. At its I/O 2023 conference, Google revealed plans to integrate the technology into its office applications, something that Microsoft has been doing for a few months already. Now, a new report claims that ChatGPT's creator, OpenAI, is looking to release a new, open-source AI model soon.

Citing an unnamed source, The Information claims that a new AI model from OpenAI is on the way, with the key highlight being that it's open-source. This would be a significant development considering that GPT-4, the large language model (LLM) powering ChatGPT, is closed-source. What this means is that with GPT-4, the public doesn't have access to the source code or the weights used to train the model, making it extremely difficult to replicate while offering OpenAI a considerable competitive advantage in this space.

However, if the latest rumor is to be believed, OpenAI is changing course and looking to release an alternative to GPT-4, one that is accessible to the public. That said, it is important to understand that this purported model will likely not pose a direct threat to GPT-4, as that would mean reducing the commercial viability of the existing LLM. Currently, OpenAI charges customers $20/month to leverage ChatGPT Plus, which is powered by GPT-4. As such, it's unlikely that the upcoming open-source model will directly compete with GPT-4.

Overall, it's not a huge surprise to hear rumors of OpenAI's open-source model, considering that other players have been making waves in this space already. Meta publicly released its open-source Large Language Model Meta AI (LLaMA) a few months ago. Even the smallest variant of the model is trained on over a trillion tokens, giving researchers and enthusiasts a large playing field to explore the underlying technology and even improve it. It's likely that OpenAI is thinking along the same lines and is willing to offer an open-source LLM to improve its overall tech and provide accessible integrations with other software.