An open-source ChatGPT rival was just launched by the Stable Diffusion team

The most recent competitor to OpenAI’s ChatGPT emerges from the creators of the renowned AI image generator, Stable Diffusion. Dubbed as StableLM, this open-source chatbot is a creation of Stability AI, aiming to make sophisticated language models more accessible to the wider public.

Stability AI has recently unveiled the alpha edition of StableLM, highlighting its comparably compact and efficient design. With a parameter count ranging from three billion to seven billion, StableLM is just 2% to 4% of the size of ChatGPT’s model, which boasts 175 billion parameters.

Similar to how Stable Diffusion serves as an image generator that’s easily extendable by third-party developers, StableLM presents an analogous open-source offering as an AI chatbot, accessible to everyone.

The most recent competitor to OpenAI’s ChatGPT emerges from the creators of the renowned AI image generator, Stable Diffusion. Dubbed as StableLM, this open-source chatbot is a creation of Stability AI, aiming to make sophisticated language models more accessible to the wider public.

Stability AI has recently unveiled the alpha edition of StableLM, highlighting its comparably compact and efficient design. With a parameter count ranging from three billion to seven billion, StableLM is just 2% to 4% of the size of ChatGPT’s model, which boasts 175 billion parameters.

Efficiency is of paramount importance to Stability AI, as it endeavors to enable StableLM’s functionality on budget-friendly systems and less potent GPUs. The alpha version of StableLM can be readily installed and operated today. Comprehensive instructions are available within the GitHub repository, accompanied by a notebook detailing its utilization on computers equipped with limited GPU capabilities.

For a straightforward trial of StableLM, the Hugging Face demo page serves as the most convenient option. Given the recent launch and the anticipated surge in demand, response times might be sluggish. Furthermore, as an alpha release, the outcomes may not match the quality of the final version.

As an illustration, when I requested assistance from StableLM in composing an apology letter for accidentally breaking someone’s phone, the AI’s response took an unexpected turn. It somehow misconstrued the situation, believing that I had presented a gift instead of causing damage to a phone.

Stability AI includes a disclaimer regarding the outcomes, given that StableLM is a pretrained Large Language Model without any supplementary fine-tuning. Unlike ChatGPT, it doesn’t utilize reinforcement learning, which means that the generated responses “may vary in quality and could potentially contain offensive language and perspectives.”

The potential of the forthcoming upgraded StableLM models to rival ChatGPT remains uncertain. Currently, it’s evident that the project is still a work in progress. This parallels the scenario encountered with another open-source contender named CollosalGPT.

However, the narrative doesn’t conclude here. Stability AI has revealed ongoing efforts to develop more expansive models, encompassing 15 billion, 30 billion, and 65 billion parameters, all aimed at enhancing the output quality. Furthermore, a future blueprint includes a 175 billion parameter model. Given the current constraints on model size, the initial strides of StableLM are promising.

The alpha version of StableLM, characterized by its open-source nature and lightweight implementation, is designed to empower developers to initiate work on applications. With significant room for expansion and enhancement, it’s certainly advisable to keep a close watch on this emerging AI chatbot.

This Post Has One Comment

Leave a Reply