Yes I Understand How LLMs Work (kemendo.com)

A recent article disputes the prevalent notion that "nobody knows how large language models (LLMs) work." The author, who has extensive experience in AI, emphasizes that the mechanisms underlying LLMs are well understood by many researchers in the field. By highlighting the mathematical foundations and operational details of these models, the article argues that LLMs are complex yet comprehensible systems built from designed algorithms and computational processes. It explains that, contrary to claims of mystery, the internal workings of LLMs—including tokenization, transformer architectures, and training methodologies—are observable and manipulable. This clarification is significant for the AI/ML community as it counters misconceptions that may hinder public understanding and acceptance of AI technologies. By framing these models as products of human ingenuity rather than enigmatic entities, the article encourages a more informed discourse about AI. It promotes ongoing research into LLMs and stresses the importance of scientific rigor in understanding and improving these systems. The author argues that while some questions about AI remain, claiming nobody understands how these technologies function detracts from genuine scientific inquiry and progress.