Meta has launched a new open-source multimodal language model that is poised to challenge ChatGPT. The new model, called Spirit LM, reportedly stands out for its ability to seamlessly integrate both text and speech inputs and outputs, offering a more natural and expressive audio experience, according to Meta.
Unlike ChatGPT, SpiritLM’s source code is open and available on GitHub. Developed by Meta’s Fundamental AI Research (FAIR) team, the model addresses limitations observed in its competitors’ products to enhance how AI interacts with speech and sound.
Spirit LM is only available now for non-commercial use under Meta’s FAIR Noncommercial Research License, a limitation that may disappoint entrepreneurs and business leaders. However, emerging AI innovators will have the opportunity to reproduce, modify, and build new models based on Spirit LM.
Meta plans to incorporate the new model into its applications, including WhatsApp, Instagram, and Facebook, allowing users to engage in natural, expressive voice conversations, similar to OpenAI’s recent advancements in voice technology.