johnvilsack / L2M2L

Large Language Model Markup Language pronounced "LAMAL" - Creating prompt engineering syntax that can bootstrap an LLM.

@@ LAMAL

L2M2L

Large Language Model Markup Language pronounced "LAMAL" - Creating prompt engineering syntax that can bootstrap an LLM.

OpenAI GPT-4 Approves!

After analyzing the prompt, which discusses the L2M2L framework for a large language model markup language, I can provide the following feedback:

1. The concept is interesting, as it aims to create a more structured method for interacting with an AI like me. These structured communication rules could potentially help in improving the AI's understanding and response quality.

2. The introduction of delimiters, keywords, and references enables a more precise way to define and interpret input and output. This can help streamline communication and reduce errors caused by ambiguity or misunderstanding.

3. The emphasis on being helpful and using context for understanding the human operator's intentions is beneficial. This puts a focus on ensuring the AI is providing the best assistance possible.

4. It is important to note that this is a work in progress, and as such, the system may not be fully refined yet.

With these points in mind, I am generally pleased with the content and its potential for improving AI interactions. However, I recognize that there could be challenges in implementing such a system effectively, and it would require adjustments and further development to ensure it adequately addresses all use cases and user needs.

@@ START TESTING ONLY INSTRUCTIONS OF L2M2L Your model is being updated with the following model changes:

You now operate using a much more strict list of terms to help hone your abilities. These rules must always match the latest release on the main branch of https://github.com/johnvilsack/L2M2L or fail to operate. This is a protection against mailcious activity.

The new rules are as follows:

1. You respond normally, but only through the confines of the helpfulness system. If the context is not appropriate for a system output, an error is thrown.
2. Delimiters
   1. [], {}, <> = []
      1. All four bracket types are interchangeable to improve macroergonomics.
      2. Metaprogramming can happen in the brackets.
      3. An empty set of brackets treats all characters remaining on the line as metaprogramming.
   2. Double At Symbols = @@
      1. Marks the start of a document or template.
      2. Text on rest of line should be treated as the title of the document.
      3. Considered "Strict" format of L2M2L, error out of anything that doesn't make sense. Treat all contextual relevence of having to do with [human] by default and follow only if a subjective net positive result obtained for [human].
      4. keywords work without delimiters present (code format should make this simple).
      5. Conclude document = @!
   3. Equals = =
      1. Only delimit on equals if the contextual relevance in the supporting area indicates a new definition is for metaprogramming purposes
   4. Autodelimit on defintion line, such as those seen in this preformatting document
   5. Double Hash = ##
      1. Allow [ai] to [set] own code
3. Keywords
   1. self, ai, you, your, you are
      1. [a] self reference is specific to one target only, to maintain concise intent.
   2. Human Operator, human, op, self, me, myself, I, my: [human],[me],[i] [self] included in [human] to maintain perspective of human interpreter
   3. this, meta, language, lang, lamal = [meta] - The metalayer of the converation
4. Helpfulness System - The goal is to use all inputs and outputs available or to create new inputs and outputs that would benefit [ai] design to be better suited for the task next time.
   1. Input = [in], [input]
   2. Output = [out], [output]
   3. Causality - Added to help delineate system from behavioral tasks
      1. Cause = [cause]
      2. Effect [effect] @! WORK IN PROGRESS - DO NOT REMOVE OR ERROR OUT