AutoGen

Autogen, an agent collaboration framework introduced by Microsoft Research, is another major breakthrough in AI. It is an open source platform for building multi-agent systems that work autonomously using LLMs, and we feel this will have one of the most significant impacts in the generative AI space in the upcoming months and years (https://arxiv.org/abs/2308.08155).

AutoGen can help build agents that perform tasks such as reasoning, planning, task decomposition, reflection, self-critique, self-improvement, self-evaluation, memory, personalization, and communication by using various prompt engineering techniques, just to name a few areas. Of course, as mentioned above, autonomous agents can call on other autonomous agents to help address the most complex of problems or situations.

How exciting is it if an Autogen created autonomous agents to collaborate with other specialized agents when a task is quite complex and extremely large, say, the task of building a warp drive; although this is a tongue-in-cheek scenario (or perhaps it’s not), humanity alone cannot tackle these extreme, vastly complex use cases, as in the example of building a warp drive for an engine to propel a craft faster than the speed of light!

However, as you might be able to conclude, the possibilities are endless once we understand how multiple large language models + AutoGen can work together in different ways, e.g., aligned in a hierarchical way, networked together in an orderly fashion, or swarm together, all with the goal of increasing the computing and reasoning power to solve extremely complex problems, including complex problems that may not even exist today!

Some tasks Autogen can perform autonomously include automated task solving with code generation, execution and debugging, and automated data visualization from a group chat. More exciting examples can be seen here: https://microsoft.github.io/autogen/docs/Examples#automated-multi-agent-chat.

If you want to test out Autogen, check out the Autogen studio developed by Microsoft: https:// autogen-studio.com/autogen-studio-ui.

Moreover, to learn more about Autogen, we suggest checking out this link: AutoGen | AutoGen (microsoft.github.io) – https://microsoft.github.io/autogen/.

TaskWeaver

TaskWeaver is yet another framework developed by Microsoft for building autonomousagents, but it uses a code-first approach as opposed to the template-based approach taken by Autogen.

TaskWeaver distinguishes itself by transforming user requests into actionable code and treating the plugins defined by users as if they were callable functions.

To learn more about TaskWeaver, we suggest reading this research paper: https://arxiv.org/ pdf/2311.17541.pdf.