As everyone tries to read the tea leaves for artificial intelligence, attempting to divine which companies will land on top and how AI will affect people’s livelihoods, one company that looks to be on the cutting edge of machine learning is LangChain Inc., which focuses on AI-powered app development.

“We’re a framework for building these apps,” Harrison Chase (pictured), founder and chief executive officer of LangChain. The company provides developers with an orchestration layer for applications that use AI. “As part of that, we have tons of integrations with all these different companies, all these different language model providers, all these different vector database companies, all these different tools.”

Chase spoke with theCUBE Research guest analyst Howie Xu at the “Supercloud 6: AI Innovators” event, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed the future of app development and how AI models can continue to grow.

Navigating the early days of LLMs

The tech industry is still in the early stages of figuring out use cases for AI, according to Chase, who noted a split between creative applications, such as ChatGPT, and more confined use cases, such as automation for customer support.

There is “a lot of experimentation and tinkering that needs to happen to figure out what the right use cases are,” Chase said, noting that less creative applications for AI have more constraints. “In creative applications … hallucinations are a feature, not a bug.”

Accuracy continues to be a major issue for large language models, with developers using Retrieval-Augmented Generation to enhance the reliability of their outputs. LangChain’s developer platform, LangSmith, is designed to monitor LLM-based applications, and provide helpful flow engineering when RAG falters.

“Where RAG fails more often than not, I would say is probably in the retrieval step …  if I ask a question and the answer I retrieve is not relevant information, then obviously I’m going to get a bad response,” Chase said. “Another huge area where RAG can often be improved enormously is on the indexing and ingestion side of things.”

Agents, intelligent AI that can perceive an environment and make corresponding decisions, still have a long way to go when it comes to tasks that require more complex reasoning.

“Flow engineering is a short-term solution for the lack of planning that the models can do intrinsically,” said Chase, emphasizing that more specific applications will require layers of AI outputs. “People are going to be building these LLM systems where you maybe have multiple LLM calls or complicated sequences of flow engineering.”

The future of AI in app development

Although the uses for AI will continue to grow as the models improve, developers will not be replaced, according to Chase. Rather, they will take on a different function as AI models become a more integral part of the process.

“Maybe [developers] take on more of a role of like a tech lead or a manager, a supervisor of these agents,” he said.

The rewind and edit functionality for LLMs that is used by Devin, an AI software engineer, and supported by LangSmith could be particularly crucial, according to Chase. It would allow developers to more easily course correct LLM results.

Despite the changes on the horizon, Chase has words of comfort for potential programmers.

“Coding’s a great way to build your problem-solving abilities,” he said. “Problem-solving abilities will never go out of style.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE Research’s coverage of the “Supercloud 6: AI Innovators” event: 

Photo: SiliconANGLE