Have been working with this and very impressed so far - it’s a step ahead of LangChain agents and seems to be receiving more attention/development than LangChain was interested in committing to agents.<p>FWIW the “group research” and “chess” examples from the notebooks folder in their repo have been the best for explaining the utility of this tech to others - the meme generator does a good job showing functions stripped down but misses a lot of the important bits
Matthew Berman has a good series on AutoGen with tutorials and demos: <a href="https://www.youtube.com/watch?v=10FCv-gCKug">https://www.youtube.com/watch?v=10FCv-gCKug</a><p>However from his examples (and his own admission) it seems that AutoGen isn't benefitting from full GPT4-level performance even tho he's pointed it directly at OpenAPI GPT4 (and other LLMs). The back and forth between the agents does not produce great results even tho similar prompts pumped directly into ChatGPT seem to give better results.<p>Anyone know whats going on?
This just reminds me: I have been wondering, if you get multiple instances of GPT-4 talking to each other, each seeded with a different personality prompt, do they have interesting conversations? I suspect it would devolve in to nonsense quickly, but I’ve never seen any chat log of two GPT instances talking. Does anyone have a reference for this? Thanks.
A question for people researching LLMs and their capabilities:<p>Is there any reason to believe that the interaction of multiple agents (using the same model) will yield some emergent property that is beyond the capabilities of the agent model?<p>I'm not working with LLMs, but my intuition is that whatever these multi agent setups come up with could also be achieved by a single agent just talking to itself, as they all are "just guessing" what the most probable next token is.
The breakthrough I've had is realizing how important it is to control the conversation between agents.<p>Just like in our work environments and in our relationships, HOW conversations occur largely determines the impact of the conversation. With or without AutoGen<p>We're building a multi-agent postgres data analytics tool. If you're building agentic software, join the conversation: <a href="https://youtu.be/4o8tymMQ5GM" rel="nofollow noreferrer">https://youtu.be/4o8tymMQ5GM</a>
Unless I'm missing something, how is this library different from prompting a single chatbot: "Write a dialog in which A, B, and C, each playing a different role, have a conversation and do something D"?
Having conversations amongst agents is it like treating each agent as your traditional nodes? Maybe in the future there would be millions of nodes(agents) conversing and maybe this is how next gen AGI will form
A bunch of single-comment related threads. Others?<p><i>AutoGen: A Multi-Agent Framework for Streamlining Task Customization</i> - <a href="https://news.ycombinator.com/item?id=37855314">https://news.ycombinator.com/item?id=37855314</a> - Oct 2023 (1 comment)<p><i>Microsoft's AutoGen – Guide to code execution by LLMs</i> - <a href="https://news.ycombinator.com/item?id=37822809">https://news.ycombinator.com/item?id=37822809</a> - Oct 2023 (1 comment)<p><i>Making memes with Autogen AI (open source LLM agent framework) [video]</i> - <a href="https://news.ycombinator.com/item?id=37750897">https://news.ycombinator.com/item?id=37750897</a> - Oct 2023 (1 comment)<p><i>AutoGen: Enabling next-generation large language model applications</i> - <a href="https://news.ycombinator.com/item?id=37647404">https://news.ycombinator.com/item?id=37647404</a> - Sept 2023 (1 comment)<p><i>AutoGen: Enabling Next-Gen GPT-X Applications</i> - <a href="https://news.ycombinator.com/item?id=37220686">https://news.ycombinator.com/item?id=37220686</a> - Aug 2023 (1 comment)
It doesn’t help you inherently solve the problem per se, but what it does allow you to do that is distinctive is keep the human and the loop that can assist the agents to solve problems.
To some degree it can also keep problems in the logic chain from snowballing, and causing the overall objective to fail because there’s invalid logic in the sequence
Are these 'safer' than using langchain-based agents that directly execute (arbitrary!) Python code? That was always my main issue with langchain