It’s like saying that you don’t need to speak a foreign language to make a deal with your host in a foreign country: just speak in your mother tongue and an agent instantly translates to Chinese or Russian what you are asking in English and signs the deal for you… Good luck!
When I was at the high school I went into the library, looked around and said: “oh my God, it will take me years to read all those classics written across centuries!“
So, who says that you don’t need a doctor to analyse a radiography because AI does it, or that you don’t need to study a programming language since AI generates code for you, he is fundamentally wrong.
There is no semantics in this book. It’s made of symbols, only syntax. The meaning was in author’s mind. Writing is like laying out the dots of a puzzle. Reading is like connecting the dots. The meant image is reconstructed in reader’s mind, according to his interpretation.
In short: humans have a small memory but may recognise you in an unpredictable situation; AI agents have a big memory but cannot identify you outside of a controlled environment.
The LLM design principles imply a limitation regarding context identification IMHO. Humans can identify and recall a conversation held a long time ago, maybe with a stranger.
But, by design, a new conversation happening in a new situation is just a new one and, without user input, it cannot be identified as the continuation of a past one, in order to provide a specific and dedicated reaction to each one of them.
@Pontifex_it Una politica sana si fonda sulla partecipazione, che non significa esprimere un voto, ma sapere chi è il proprio rappresentante e come opera, attraverso un canale di comunicazione istituzionalizzato ed efficiente con lui, sulla base della vicinanza territoriale.