The Allen Institute for Artificial Intelligence (AI2), AMD and Large Unified Modern Infrastructure cooperated to plan a large-scale research-based Open Language Model (OLMo) in 2024.
Image source: pixabay
Note: Including PaLM 2 and GPT-4, most of the models used in current generative AI are hidden behind the API, and researchers cannot check the internal information exactly. So more and more communities are trying to create open source AI models.
Hanna Hajishirzi, Senior Director of NLP Research at AI2, said: “The research and technical communities need to push AI further by having access to open language models. With OLMo, we are working to narrow public and private research capacity and knowledge gap”.
AI2 sees OLMo not just as a model, but as an open platform that allows the research community to use all components created by AI2 and improve them through open source.
Everything from AI2’s demos, training datasets, and APIs for OLMo will be publicly available, Hajishirzi said.
