Ep 28: LangChain CEO Harrison Chase on the Current State of Eval and Agents and The LLM Apps that Will Define 2024

Current Trends

Harrison Chase, CEO of LangChain, highlights the evolving landscape of AI applications, drawing parallels with past technological shifts like the iPhone and the Internet. He notes that while AI's impact is rapidly unfolding, the most transformative applications may still be on the horizon, similar to how Uber and Facebook emerged years after foundational tech developments 1. LangChain, as a framework, plays a pivotal role in this evolution by providing an orchestration layer for building diverse LLM applications, from chatbots to data retrieval systems 2.

LangChain started as a framework for building LLM applications, which itself is very vague and horizontal because LLMs themselves are very vague and horizontal.

--- Harrison Chase

The framework's versatility is evident in its ability to connect language models to external data and computation sources, facilitating a wide range of use cases 2.

   

Evolving Use Cases

The use cases for AI applications are becoming increasingly sophisticated, moving beyond basic chat functionalities to more complex solutions. Harrison Chase observes a shift towards advanced applications like state machine-based chatbots and longer-running jobs that require multiple LLM calls 3. These innovations are particularly evident in internal enterprise projects, where risk-taking is more feasible, allowing for the development of platforms that integrate language models with proprietary data and APIs 4.

I think all the most interesting work right now is in UX for these AI applications. I think that's where there's a ton of innovation to be done.

--- Harrison Chase

This evolution signifies a growing maturity in AI applications, as developers explore new ways to leverage language models for more intricate tasks 4.

   

Future Predictions

Looking ahead, Harrison Chase envisions a future where AI applications become even more integrated and impactful. He emphasizes the early stage of the AI space, suggesting that many opportunities remain untapped 5. As AI models improve, some current practices, like managing conversation history, may become obsolete, while the need for retrieval and state machine frameworks will persist 6.

The only thing I have confidence in is that the space is early. I really do think there's a ton of opportunity to build not only applications using this technology.

--- Harrison Chase

Chase's insights underscore the potential for AI to revolutionize various sectors, driven by continuous advancements in technology and user experience 5.