When one considers LLM versions of AI one thinks of asking a question and getting an answer. I recall back some fifty five years when at MIT I went to scan through journals looking for some info and to my surprise the journals disappeared. The head of the library informed me that they were now all on microfiche, a convenient was for the library to store them. She also lectured me on the efficiency to me rendered by getting exactly what I asked for!
I had a bit of a fit. I explained that by having the whole journal I had the opportunity to randomly come across something I would never have asked for. Those random occurrences all too often led to new insights and expansions of what I was looking for. Rigid and delimited searches, such as what LLM AI preforms fails to provide those fortuitous occurrences. I would say that half of what I wrote was based on such random reads, something I was not looking for but which bounced to the fore.
LLM AI are for the most part rigid. You ask a question and the algorithm provides an answer based upon your question and the data it has been fed. The loss of random insights is grossly absent.
Posts
