Inspiration

We all know that when people collaborate, we can achieve things that no single one of us could do on our own. Is the same true for LLMs?

What it does

I've built a prototype with two LLMs (Bard & GPT-4) collaborating. Only Bard knows about current events, but GPT-4 is far better at formatting data and writing reports. By getting the two to work together, we can produce concise reports on news happening today.

How we built it

Used prompt engineering to get completions from Bard that could be easily digested by ChatGPT, then piped those to ChatGPT with additional prompt engineering to instruct it to read and format data in a certain way, as well as add its own context.

What we learned

That it's possible! Spoiler alert! But moreover, this is likely to be a significant source of collaboration in the future.

What's next for LLM Collaboration

Right now, collaboration is fairly manual. What if you could directly connect the two, and engineer prompts from both to let them build off each other's ideas, just like people do?

Built With

  • bard
  • gpt-4
Share this project:

Updates