What’s next for agentic AI? The LangChain founder looks at the surrounding factors

Agent AI is the latest big trend in generative AI, but what comes next?

While full artificial general intelligence (AGI) is likely still at some point in the future, there may be a more intermediate step with an approach known as ambient agents.

langshenLangChain, a pioneer in the field of artificial intelligence, introduced the term “ambient agents” on January 14. The technology developed by LangChain includes the open source LangChain framework that enables organizations to link different large language models (LLMs) together to produce an outcome. LangChain Inc. 24 million dollars In funding in February 2024. The company also has a series of commercial products including LangSmith for LLM Ops.

Using a traditional AI interface, users typically interact with the LLM via text prompts to initiate the action. Agentic AI generally refers to LLM-enabled systems that take actions on behalf of the user. The concept of surrounding factors takes this model a step further.

What are the surrounding factors?

Ambient agents are artificial intelligence systems that operate in the background, continuously monitoring event streams and then acting when appropriate due to triggers, according to pre-defined instructions and user intent.

While the term environmental factors is new, it is a concept Ambient intelligencewhere AI is always listening, is not. Amazon refers to its Alexa personal assistant technology as enabling ambient intelligence.

The goal of ambient agents is to automate repetitive tasks and extend user capabilities by running multiple agents continuously, rather than a human user having to call them up and interact with each one of them. This allows the user to focus on higher-level tasks while the agents handle the routine work.

To help prove the concept and advance its perimeter agents, LangChain developed a series of initial use cases, one monitoring emails and one for social media, to help users manage and respond when needed.

“I think agents in general are powerful, exciting, and cool,” Harrison Chase, co-founder and CEO of LangChain, told VentureBeat. “Ambient agents are much more powerful if there’s a group of them doing things in the background, you can just expand your scope more.”

The technology leverages several open source solutions, and LangChain has not yet said how much it will charge for use of any new tools.

How environmental factors improve the usability of artificial intelligence

Like many great technological innovations, the original motivation for Ambient Factors was not to create a new paradigm, but to solve a real problem.

For Chase, the problem is familiar to many of us: email inbox overload. Chase began his journey to create surroundings to solve email challenges. Six months ago he started building a perimeter proxy for his own email.

Chase explained that the email assistant categorizes his emails, and handles the sorting process automatically. He no longer has to manually sort through his inbox, as the agent takes care of it. Through his use of the agent’s inbox over an extended period, Chase was able to refine and improve his capabilities. He pointed out that the beginning was not ideal, but by using it regularly and treating pain points, he was able to improve the worker’s performance.

To be clear, Email Assistant is not some sort of simplistic, rules-based system for sorting email. It’s a system that actually understands his email and helps him decide how to manage it.

Perimeter proxy architecture for email assistant use case

The architecture of Chase’s email assistant is very complex, involving multiple components and language models.

“It starts with a sorting step that’s kind of like an LLM, a very complex vector, and some few examples that are semantically retrieved from the vector database,” Chase explained. “Then, if it is determined that he should attempt a response, he goes to the drafting agent.”

Chase also explained that the drafting agent has access to additional tools, including a dedicated sub-agent for interacting with the calendar:

“I have an agent specifically to interact with the calendar, because the LLM is actually not good at dates,” Chase said. “So I had to have a dedicated agent just to interact with the calendar.”

After creating a draft response, Chase said there is an additional LLM call that rewrites the response to ensure the correct tone and format.

“I found the MBA trying to call up all these tools and create an email and then also write in the right tone was really difficult, so I have a clear tone step,” Chase said.

Agent inbox as a means of controlling and monitoring agents

A key part of the ambient agent experience is control and visibility into what agents do.

Chase noted that in the initial implementation, he only had agents message him via Slack, but that quickly became unwieldy.

Instead, LangChain designed a new user interface, the Agent Inbox, specifically for interacting with surrounding agents.

The system displays all open lines of communication between users and agents and makes it easy to track pending actions.

How to build an ambient proxy

LangChain is primarily a tool for developers and will be a tool to help create and deploy surrounding agents now as well.

Any developer can use open source LangChain technology to create a perimeter agent, although additional tools can simplify the process. Chase explained that the agent inbox he created is in some ways a view on top of the LangGraph platform. LangGraph is an open source framework for build agents that provides the infrastructure to run long-running background jobs.

Furthermore, LangChain uses its own LangSmith trading platform which provides monitoring and evaluation capabilities for agents. This helps developers put agents into production with the necessary monitoring and evaluation tools to ensure they perform as expected.

Surrounding factors: A step towards the use of generalized intelligence

Chase is optimistic that the concept of ambient factors will attract developers in the coming months and years.

Surrounding factors provide the potential for greater autonomy for AI, enabling it to monitor the flow of events to be able to take intelligent actions. Chase still expects that there will be a need to keep humans in the loop as part of the Ambient Agent experiment. Humans only need to confirm and validate actions, rather than knowing what to do.

“I think it’s a step toward harnessing and using more general intelligence,” Chase said.

True artificial general intelligence is likely to come from improvements in reasoning models, Chase noted. However, the best use of models is where the concept of surrounding factors will bring value.

“There is still a lot of work to be done to take advantage of the models, even after they become really smart,” Chase said. “I think the ambient agent’s style of interacting with them will definitely be an opening to using this general form of intelligence.”

An open source version of the email assistant is currently available. LangChain is launching a new social media agent today and will provide an open source version of the agent inbox on Thursday, January 16.