Databricks has released a dataset to support the training of AI models. The software company is promoting the development of AI to acquire new customers. The data consists of 15,000 questions answered by 5,000 company employees from 40 countries. CEO Ali Ghodsi has ensured the quality of the dataset, though he admits it is not perfect. He hopes competitors to OpenAI’s successful ChatGPT will become available.
This promotion of AI training could be a good marketing strategy for Databricks. The American company offers a wide range of services, including lakehouse products and AI. Unlike GPT-4, Databricks wants to use LLMs for specific purposes. OpenAI keeps the dataset on which GPT-4 is based secret, while Databricks is trying to support a different approach for customers looking to deploy AI in a more focused way. Databricks is demonstrating what is possible with LLMs trained on their own datasets with their research model Dolly.