What happens when two large language models (LLMs) converse?

What is the outcome? Are there any science fiction books that explore this concept?

Or, you could set up a test and tell us!

To avoid unnecessary duplication of effort, similar to the well-known rediscovery of calculus in 1994 [1], I will leverage existing resources.

[1] [Rediscovery of calculus in 1994 - PubMed](https://pubmed.ncbi.nlm.nih.gov/8137688/)

Hm. Not on the order of 'reinventing calculus'. More like, piping the output of two LLM command lines to each other, letting them run for a couple of hours and examining the results.

Reply on news.ycombinator.com