The creation of DBRX adds momentum to the movement challenging the secretive approach of leading companies in the AI field. While OpenAI and Google keep their large language models closely held, Databricks and Meta have released their models for others to use, arguing that it will spur innovation. Databricks also plans to be transparent about the work involved in creating its open-source model, a move welcomed by AI researchers who have called for greater transparency in the field.
Key takeaways:
- Databricks, a data science and AI company, has developed a powerful artificial intelligence language model called DBRX, which has outperformed all other open source models in various benchmarks.
- The company spent months and about $10 million to train DBRX, which is similar in design to OpenAI’s ChatGPT, and plans to release it under an open source license.
- DBRX also came close to matching the performance of GPT-4, OpenAI’s closed model that powers ChatGPT and is considered the pinnacle of machine intelligence.
- Databricks aims to help companies in finance, medicine, and other industries use AI technology on their own data, and will customize DBRX for a customer or build a bespoke one tailored to their business from scratch.