The nonprofit Allen Institute for AI Research (AI2) is developing the Open Language Model (OLMo) in collaboration with AMD, the Large Unified Modern Infrastructure consortium, Surge AI, and MosaicML. Scheduled for release in 2024, OLMo aims to offer the AI research community an open language model optimized for scientific and academic applications.
According to Hanna Hajishirzi, Senior Director of NLP Research at AI2, OLMo seeks to "close the gap between public and private research capabilities and knowledge by building a competitive language model." AI2 envisions OLMo as a platform rather than just a model, allowing researchers to utilize or improve upon each component AI2 creates. All aspects of OLMo will be openly available, including a public demo, training dataset, and API, with minimal exceptions under suitable licensing.
OLMo's focus is to better understand and leverage textbooks and academic papers, distinguishing it from other open-source models. AI2's background in academia and its development of tools like Semantic Scholar will help make OLMo uniquely suited for scientific and academic applications.
To address ethical and legal concerns surrounding generative AI, the OLMo team plans to work with AI2's legal department and outside experts at various checkpoints in the model-building process to reassess privacy and intellectual property rights issues. AI2 will use a combination of licensing, model design, and selective access to underlying components to balance scientific benefits and the risk of harmful use. OLMo also has an ethics review committee with internal and external advisors to provide feedback during the model creation process.
AI2 is inviting collaborators to contribute to and critique the model development process.