Hutter-Prize for Prompts

The aim of the Hutter Prize is to compress the first 1GB of Wikipedia to the smallest possible size. From the AIXI standpoint, compression is equal to AI, and if we can compress this to the ideal size (75MB according to Shannon’s lower estimate), then the compression algorithm is equivalent to AIXI.

However, all the winning solutions so far are based on arithmetic encoding, context mixing. These solutions hold little relevance in mainstream AGI research.

Current day LLMs are really powerful models of our world.And it is highly possible that most LLMs are trained on data that consist of the first 1GB of Wikipedia. Therefore, with appropriate prompting, it should be possible to extract most or all of this data from the LLMs.

I am looking for ideas/​papers which would help me validate whether it is possible to extract pieces of wikipedia using prompts with any publicly available LLMs.

update: i have recently learnt regarding gpt-4′s compression abilities via prompting, i am very keen in testing this using this method. If anyone is willing to work with me (as i do not have access to gpt-4) it would be great.