Exploring Multi-Agent AI and AutoGen with Chi Wang

In this episode, I’m joined by Chi Wang, a principal researcher at Microsoft and the creator of AutoGen, a open-source framework that allows developers to combine LLMs, tools, and human input to build multi-agent AI systems.
By enabling AI agents to collaborate, learn from each other, and contribute their unique skills, AutoGen is unlocking a new frontier of AI capabilities. It's quickly gained traction among both academics and enterprises and is currently powering a wide range of use cases, including synthetic data generation, code generation, and pharamaceutical data science.
In our conversation, Chi breaks down the core concepts behind multi-agent AI, the pros and cons of multi-agent architectures, and the real-world use cases enabled by AutoGen. He also shares some of the open research challenges he's tackling, his perspective on the future of AI, and what excites him most about where the field is headed.
Read takeaways from our conversation here: foundationcapital.com/the-pro...
(00:00) Intro
(01:27) Chi's background and early interest in AI
(05:42) Defining agents and their core capabilities
(08:13) Pros and cons of multi-agent systems
(11:23) Multi-agent architectures and the "Society of Mind" theory
(14:36) Real-world use cases enabled by multi-agent systems
(16:45) The backstory and genesis of AutoGen
(19:43) How AutoGen's architecture leverages language models, tools, and human input
(23:13) More examples of AutoGen's diverse applications
(29:23) How AutoGen is being used in enterprises and production
(32:42) Advice for AI builders focusing on the enterprise
(40:24) What's next for AutoGen and open research challenges
(47:17) Resource recommendations for AI builders
(48:09) What excites Chi most about the future of AI

[@filippopirri3757]

I'm curious to hear about what is the best approach to build multi-agents workflows in AutoGen. In particular, do you: - set up and debug every single agent, and then put them together in a group chat - or rather start building within the group chat already?

[@pin65371]

I am still learning all of this, but it seems to me that the best approach is to start with multiple agents. Each agent should perform a very specific task exceptionally well and log everything. By doing so, you can build large datasets on how each agent operates and fine-tune them accordingly. Over time, you can potentially combine agents to improve efficiency. Initially, however, it is important to observe how they handle specific tasks. I could be wrong, but it appears that if you can precisely define what the LLM is doing, you can prevent hallucinations. This approach might involve higher upfront costs, but in the long term, it could lead to faster progress, as you wouldn't be trying to solve problems with agents that are stretched too thin. The advantage here is that, unlike starting a company where having too many employees significantly increases costs, using multiple agents does not lead to a substantial cost increase. While compute costs money, it is not as expensive as hiring additional human employees.

[@michaelpaine8849]

Thank you, great tooling. I can't wait to see it grow

[@fintech1378]

what are benchmarking and evaluation tools for agentic workflow

[@FranAbenza]

GAIA benchmark