Stable Diffusion creators introduce GPT-4 alternative
The language model, which is made available as open source, is called StableLM and was developed by the company that also released the AI image generator Stable Diffusion.
Stability AI was one of the first companies to make a name for itself in the AI world. However, it then took its time to present its large language model, which is now competing with OpenAI’s GPT-4, Meta’s Llama and Google’s Lamda.
StableLM stands for a set of open source Large Language Models (LLM). Developers can now open them GitHub use freely. The AI is powered by an open-source dataset called Pile, which includes information from a variety of sources, including Wikipedia, Stack Exchange, and PubMed.
According to Stability AI, StableLM models currently have between three and seven billion parameters, later it will be 15 to 65 billion. For comparison: GPT-3, the predecessor of GPT-4, was already trained with 175 billion parameters.
With StableLM, Stability AI continues on the path it has already trodden with Stable Diffusion: making AI tools more accessible to the general public. The company made its text-to-image AI available in a variety of ways, allowing developers to play around with the tool and build various integrations. At the moment a trial version of StableLM is up hugging face available.
However, initial examples show that AI is still in its infancy. She did not sufficiently comply with the request to create a peanut butter sandwich recipe, reports a Twitter user:in.
There’s still a long way to go, admitted Emad Mostaque, CEO of Stability AI. On Twitter he explained that things will get better in the next three to six months “and great for years to come.”
Loud gizmodo However, the company has been struggling with money worries lately after spending so much on developing its AI projects.
