Databricks to democratize the magic of ChatGPT with open LLM models
Databricks, a leading data lakehouse platform provider to government agencies, has demonstrated an approach to democratizing the magic of ChatGPT by using existing, open-source Large Language Models (LLMs).
In a post on the company’s blog, they explain how they trained an open-source LLM in 30 minutes on one machine, using high-quality training data. Their model contained only 6 billion parameters, compared to the 175 billion parameters used by GPT-3.
Databricks has open-sourced the code for their model, nicknamed Dolly, to show how customers (including federal agencies) can experiment with recreating the model on the Databricks platform. They expect that this approach will turn LLMs into a tool that any company or organization can afford, leading to the increased democratization of data.