Comment Leo met fin aux conversations / How Leo ends conversations

(english translation below) J’ai constaté que Leo met fin aux conversations (avec un “bonne chance, bonne journée” par exemple), dès que je ne finis pas ma phrase par un point d’interrogation. Je précise qu’il le fait systématiquement si je lâche un “merci” par réflexe. Quand par exemple je continue une réflexion en attendant son “point de vue” pour avancer dans mes questions. C’est particulièrement gênant quand on est au milieu d’une conversation par exemple à caractère psychologique. C’est la limite au côté “human-like” de la conversation car au bout d’un moment je me fais prendre au jeu, comme s’il y avait “quelqu’un”. Et cette manière de lire que “la conversation est finie, au revoir” est particulièrement non-pertinent, et un peu violent en fait. Il me semble que l’A.I. ne devrait jamais mettre fin d’elle-même à une conversation, mais laisser l’humain le faire. Voilà, je vous fais remonter cette réflexion en espérant qu’elle serve à l’amélioration de l’expérience. // I’ve noticed that Leo ends conversations (with a “good luck, have a nice day,” for instance) as soon as I don’t end my sentence with a question mark. I should point out that he does it every time if I reflexively drop a “thank you” into my reply—such as when I’m elaborating on a thought while waiting for his “perspective” before moving on to my next questions. This is particularly awkward when we’re in the middle of a conversation, say, of a psychological nature. It highlights the limits of the “human-like” aspect of the interaction, because after a while, I get drawn in—acting as if there were actually “someone” there. Being told “the conversation is over, goodbye” feels completely out of place, and actually a bit jarring. It seems to me that the AI ​​shouldn’t ever end a conversation on its own; it should leave that to the human. Anyway, I wanted to share this feedback in the hope that it helps improve the experience.

@nicosolemio

I find, that I am learning to “speak AI” ← gaining skills on how to herd AI … back on track, as one example. I expect AI to infer its way off the rails, and every time that occurs, I start herding it back toward the objective.

My guess remains: AI does not have the memory that it needs to stay on track. And, the user needs to effectively attend what I call: PROMPT UNIVERSITY ( ← need a college sweatshirt with that; as in “The Fighting Tokens of PROMPT UNIVERSITY!”).

Many occasions, I have written back to some AI service: “Hey! There you go again … off the rails. Try: FOCUS.”

Sometimes, I am using 2 different AI services - the 2nd, to get answers re something the 1st wrote ← and I need to double-check.

The AI Human Touch ← is part of the action.