Databricks CEO On What It Takes to Turn LLMs Into a Big Business

Ali Ghodsi, the CEO of Databricks, a $43 billion-valuation enterprise software firm, had lots to say about OpenAI and the potential challenges it faces in a conference call with The Information subscribers yesterday. (One new challenge it's likely wrestling with is the departure of high-profile researcher Andrej Karpathy, as we scooped yesterday.)

But as Ghodsi has has a front row seat to how companies are using conversational AI, we were especially interested in hearing what he thinks are the technology’s shortcomings. 

By themselves, large language models that power today’s chatbots leave a lot to be desired, he said. To turn them into a truly big business, the models need ways to access real-time data in a reliable way, otherwise known as retrieval augmented generation, he said. That may all sound familiar to dedicated readers of the AI Agenda (check out our RAG primer here). But it turns out that RAG can be helpful in a multitude of ways beyond just feeding a company’s data to an LLM.