Is Large Language Models Potential Really Exhausted ? Insights from Meta-Analysis and Field Observations

Abstract

Current limitations both of LLMs development and implementation are discussed from a unique perspective : crossing the technical expertise and field research from social sciences, present paper demonstrates how the problems faced by models transcends pure technicality and echoes on user’s level. Scaling and architectural limitations are synthetically formulated based on meta-analysis of existing studies and reports while the the potential to master them is explored through the development of systems which LLMs become the core element of. Authors provide their vision of the most plausible scenario for LLMs further development, taking in consideration current economic and institutional realities.