Yesterday, Stability AI unveiled StableLM, an open source AI language model family designed to offer an alternative to ChatGPT.
Stability AI, a London-based firm, positions itself as an open source rival to OpenAI. With the alpha release of StableLM on GitHub, available in 3 billion and 7 billion parameter model sizes, the company is fostering innovation and collaboration under the Creative Commons BY-SA-4.0 license.
Language models are becoming central to the digital economy, and Stability AI is committed to creating transparent, accessible, and supportive AI technology. StableLM operates similarly to GPT-4 (the large language model powering ChatGPT) by predicting the next token in a sequence. Notably, StableLM achieves performance comparable to GPT-3 with significantly fewer parameters (7 billion vs. GPT-3’s 175 billion), which means smaller, more efficient models, ones suitable for local devices like smartphones and laptops.
StableLM’s “surprisingly high performance” is attributed to training on an experimental dataset three times larger than The Pile, an open source dataset. Larger-parameter versions of StableLM are anticipated to offer even greater capabilities.
Last year, Stability AI’s open source launch of Stable Diffusion accelerated advancements in image-synthesis technology while generating backlash from artists and corporations. Their move into language models could produce similar results.
The 7 billion parameter StableLM base model is available for testing on Hugging Face and Replicate. Hugging Face also hosts a dialog-tuned version that resembles ChatGPT’s conversation format.
Stability AI’s introduction of StableLM is an important step in the democratization of LLMs. It will help developers innovate a new generation of conversational AI tools (chatbots), customized models, private secure feedback loops, and better autonomous agents.
If you want to learn more about what open source LLMs may mean for your business, please check out our free online course, Generative AI for Executives, now with new sections about autonomous agents and security and privacy issues.
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it.