Home / Software / Apps / Large language models get private data from LlamaIndex

Large language models get private data from LlamaIndex

Last fall, former Uber research scientist Jerry Liu found “limitations” in OpenAI’s GPT-3 text-generating AI model, the predecessor to GPT-4. LlamaIndex, his open-source project, unlocks the capabilities and use cases of large language models (LLMs) like GPT-3 and GPT-4.

“LLMs offer incredible capabilities for knowledge extraction and reasoning—they can perform question-answering, summarization, insight extraction, and even sequential decision making with an external environment,” Liu told in an email interview. LLMs have limits.

Liu partnered with Uber colleague Simon Suo to turn LlamaIndex into a company after 200,000 monthly downloads. Today, LlamaIndex provides a framework to help developers use LLMs on personal or organizational data.

Liu said LlamaIndex helps developers manage LLM data. “Our toolkit has the most depth and is easy to integrate with other developer tools.”

Developers can connect LLMs to PDFs, PowerPoints, Notion, Slack, Postgres, and MongoDB using the LlamaIndex framework. The framework has connectors for data sources and formats and ways to structure data for LLMs.

Developers can use LlamaIndex’s data retrieval and query interface to get “context and knowledge-augmented” output from any LLM input prompt.

“Other LLM application frameworks offer basic building blocks for LLM applications and agents,” Liu said. LlamaIndex connects your data sources to LLMs and has extensive tools for data ingestion, management, indexing, and retrieval for LLM applications.

In a recent seed funding round, investors pledged $8.5 million to LlamaIndex to augment LLMs. Angel investors Jack Altman, Lenny Rachitsky, and Charles Xie joined Greylock.

What will LlamaIndex buy? Liu plans to build a “enterprise solution” on top of the open source LlamaIndex project later this year. Customers can index “domain-specific” data and parse and transport large amounts of data using “protection-grade” data connectors.

Liu said LlamaIndex is not tied to a specific technology so it can be used with LLMs as technology evolves. The AI industry is moving so fast that initial stacks will likely change in the next few months.

About Chambers

Check Also

The Air Force has abandoned its attempt to install a directed-energy weapon on a fighter jet, marking another failure for airborne lasers

The U.S. military’s most recent endeavor to create an airborne laser weapon, designed to safeguard …