Turning to a friend or colleague can make difficult issues easier to resolve. It now appears that AI chatbot collaboration can make them more effective.
I played this week with AutoGen, an open source software framework for AI agent collaboration developed by researchers at Microsoft and academics at Pennsylvania State University, the University of Washington, and the University of Xidian in China. The software leverages OpenAI’s large GPT-4 language model to allow you to create multiple AI agents with different personas, roles, and goals that can be asked to solve specific problems.
To test the idea of AI collaboration, I asked two AI agents to work together on a plan for how to write about AI collaboration.
By modifying AutoGen’s code, I created a “reporter” and an “editor” who discussed writing about AI agent collaboration. After speaking about the importance of “showing how industries like healthcare, transportation, retail, etc. use multi-agent AI,” both men agreed that the proposed article should delve into the “ethical dilemmas” posed by the technology.
It’s too early to write much about any of these suggested topics: the concept of multi-agent collaboration in AI is largely in the research stage. But the experiment demonstrated a strategy capable of amplifying the power of AI chatbots.
Large language models like those behind ChatGPT often run into mathematical problems because they operate by providing statistically plausible text rather than rigorous logical reasoning. In a paper presented at an academic workshop in May, the researchers behind AutoGen show that AI agent collaboration can alleviate this weakness.
They found that two to four agents working together could solve fifth-grade math problems more reliably than a single agent. In their tests, the teams were also able to solve chess problems by discussing them, and they were able to analyze and refine computer code by discussing them with each other.
Others have shown similar benefits when several different AI models, even those offered by competing companies, come together. In a project presented at the same workshop at a major AI conference called ICLR, a group from MIT and Google brought OpenAI’s ChatGPT and Google’s Bard to work together by discussing and debating problems. They found that the duo was more likely to converge on a correct solution to problems together than when the robots worked alone. Another recent paper by researchers at UC Berkeley and the University of Michigan showed that having one AI agent review and critique the work of another could allow the supervisory robot to upgrade the other agent’s code, thereby enhancing its ability to use a computer’s web browser.
LLM teams can also be encouraged to behave in surprisingly humane ways. A group from Google, Zhejiang University in China and the National University of Singapore found that assigning distinct personality traits to AI agents, such as “easy-going” or “overconfident” , can refine their collaborative performance, whether in a positive or negative way. .
And a recent article in The Economist summarizes several multi-agent projects, including one commissioned by the Pentagon’s Defense Advanced Research Projects Agency. In this experiment, a team of AI agents was tasked with searching for bombs hidden in a maze of virtual rooms. While the multi-AI team was more effective at finding the imaginary bombs than a single agent, the researchers also found that the group spontaneously developed an internal hierarchy. One agent ended up giving orders to the others as they carried out their mission.
Graham Neubig, an associate professor at Carnegie Mellon University who organized the ICRL workshop, is experimenting with multi-agent collaboration for coding. He says the collaborative approach can be powerful, but it can also lead to new types of errors because it adds more complexity. “It’s possible that multi-agent systems are the way forward, but it’s not inevitable,” says Neubig.
People are already adapting the open source AutoGen framework in interesting ways, such as creating simulated writers’ rooms to generate fiction ideas, and a virtual “business-in-a-box” with agents taking on different roles at the same time. within the company. It may not be long before the mission proposed by my AI agents is drafted.
