Large Language Models (LLMs) have been riding high, draining private equity and VC coffers as well as the cash balances of Search, OS and Social media monopolies. All browsers have now incorporated “Artificial Intelligence,” as demonstrated by the fact that you can ask them anything and you will get an answer. Most technologists ask if that is not AI, what is?
As they grunted a few hundred thousand years ago in the African savannah in an attempt to communicate, little did they know that the Silicon kings will create intelligence from their uttering much later. Once they started to record the instances of their communications, language began to grow. They captured and stored history, made flowery poems, created emotions through literature, and collaborated in scientific arenas. Language has been critical not only to human development but also destruction through wars and hatred. More recently, great politicians have been spewing garbage into the airwaves that makes one wonder if language is useful.
Language, however, is just a tool. What it has captured last few hundred thousand years is history, most of which could be wrong. Humans’ use of language seldom incorporated the future except for some fraudulent fortune tellers and religious gurus and they have always been wrong. Over half of scientific articles even now cannot be replicated. So, how is a technology that utilizes raw materials that are historical and likely wrong, get intelligent?
As a rule, models are never intelligent. And, Large Language Models less so. Those chasing “Artificial Intelligence,” using history may be in for a big surprise.