Two NLP transformer AI are trained on a curated dataset of programming language theory etc. An artificial environment is set up with the initial condition that they are experts in the field of Python. They are engaged with each other in a collaborative discussion on creating a new programming language for training and running NLP AI algorithms. They eventually begin speaking in shorthand. At this point another AI observes and instructs them to translate the language into a human understandable language. Essentially providing humans with an instruction set to create the basis for a new programming language API.
This is a thought experiment but I am curious if it would be possible theoretically or if this has been tried and failed in other domains or on other AI neural network paradigms? Circular logic, etc. I am beginning to study Machine Learning and am wildly curious. Please be kind. I’m working my way through and making mistakes like everyone else. I plan on researching further as I continue learning. Some article references would be appreciated but I thought it best to start here? Thank you.