What's next for AI agents ft. LangChain's Harrison Chase

Harrison Chase, founder of LangChain, speaks at Sequoia Capital's AI Ascent about what's next for agents in AI and the evolution of using language model to interact with the external world. Harrison identifies three critical areas of development for the next generation of agents that would make them production-ready and more impactful in the real world: planning, user experience and memory.
#AI #AIAscent #Sequoia #Startup #Founder #entrepreneur

[@samarbid13]

4:02 Talking UX, as an X VFX artist, think extracting Blender's node-based engine or litegraph, but for LangChain. Transforming AI's UX into a visual symphony of nodes. it will be a revolution in AI interaction!
This will give both visualization capabilities and much more flexibility to chain nodes and play with them.
The current presentation is suggesting the AfterEffect way of solving the problem and it will be limited. VFX industry solved this long time ago with Node based editors and we should bring this in the AI world!

[@Ranovin]

You’ll be happy to find out that I have been working on an open source implementation of this exact idea for over a year now!
Still trying to get the word out.
Hard to explain in a TH-cam comment.
The results are already pretty incredible.

[@aesuna2565]

depends on the use case. VFX uses that because they are pro users. not all users will need something that deep

[@aymn-.-3398]

@@Ranovin Sounds interesting, can you elaborate a bit? I'd love to learn more about what you are working on and maybe contribute.

[@FranAbenza]

Building a state machine of thought is so valuable.
Flow engeneering can take you far. Waiting to see a flow that actually fine tune the model on the go as part of the. long term task

[@sillystuff6247]

Thanks for these bite sized chunks of your conference.

[@paveltarashkevich8387]

First slide pic is taken from Lilian Weng great blogpost, should be referenced

[@AdvantestInc]

Balancing autonomy and human intervention in AI agents is crucial. How will this shape future UX designs?

[@emreon3160]

Glad there is progress and this is shared, at the expense of Gemini getting smarter from youtube content.

[@XShollaj]

Harrison is an amazing CEO - He is always on reddit/twitter and any platform engaging in real time with all developers and always looking for ways to improve. I definitely think LangChain currently is bloated and inefficient at times, but I like the direction and rooting for LangChain

[@asatorftw]

100% agree. He is really active and engaged. Definitely rooting for the Langchain Teams success.

[@FranAbenza]

what are the ergonomics of augmented reasoning?

[@kevinamiri909]

It is a buzzword, for promoting a product that has no value.

[@FranAbenza]

@@kevinamiri909 what product? Langchain?

[@tarundevnandan2327]

When I see Andrew NG in the thumbnail, I instantly click on it, I at the moment is kind of a fanboy of him

[@Kevin_Knights]

We all are! Specially if you got started in the AI/ML field thanks to his Coursera courses.

[@VR_Wizard]

Same here. We should make a group chat a kind of fanboy group sharing resources. If you interested let me now and I reach out to you.

[@mailuser41]

@@VR_Wizard please add me in

[@Nifty-Stuff]

Could multiple LLMs be setup as Agents? Why hasn't anybody developed a system/app that takes the API's from the top LLMs, created agents for each, and then have these agents all work together to brainstorm, debate, review, and solve problems? I often get 4 different answers from 4 LLMs, so why not have them all setup as agents "in one room" working together to come up with the "best" solution. I can't find anybody that's tried this... why not? Wouldn't having the "top minds" (LLMs) working together produce better results?

[@therealrichot]

See AutoGen and CrewAI. Those may be close to what you’re envisioning.

[@user-cr6uz2pb5m]

There is. I am basically working like that for a year now. It will be just way more efficient with agents.

[@PeterResponsible]

this is a common thing, many Open Source products leverage a similar approach. You often see calling one LLM for task A and another for task B. You can actually see this just in the Chat GPT interface - you use the smart model to chat with but it's a dumber and faster model that comes up with the appropriate name for your chat after the first question and response are returned.

[@user-eq1wo9xy4q]

currently working on coze, should definitely check that out

[@louiepecan]

First📠💯