I find the first two responses "Perfect!..." and "Based on..." intensely annoying. They add nothing to the discussion and act as noise. This article crystalises a problem with AI: words are cheap to generate and expensive to read (just like spam). The AI should just do as it's asked and succinctly generate the poll. If asked "Why did you make these suggestions", it should explain why it chose those restaurants.
Edit: I'd also add that done "right" the AI should be clearly labelled as such. Its chat handle should include something like "AI", so the other chat users know exactly what they are dealing with, without having to know what "TripJam" or "TJ" is.
Unless it's a reasoning model with access to its own reasoning trace, any questions of "why did you choose X" will result in a rationalization, not an explanation. The model simply doesn't know why it made the suggestion otherwise.
yeh I type in Paris, now I get a paragraph to read with random suggestions with no context to verify if these are good for me or not. If User has to figure out and make choices and tell AI to plan the trip then what is the point. It is easy for me to see nice UI with reviews, image grid and make choice based on this type of data. Why is the purpose of back and forth discussion with AI ?
I think a nice thing about a group chat is that the context and preferences are from users talking amongst each other instead of a long back and forth with AI
10 comments
[ 2.3 ms ] story [ 29.9 ms ] threadEdit: I'd also add that done "right" the AI should be clearly labelled as such. Its chat handle should include something like "AI", so the other chat users know exactly what they are dealing with, without having to know what "TripJam" or "TJ" is.
I have a sports betting group chat "Degen Bets" that I made a funny image for the thumbnail based on our fantasy football teams.
In a temporary thread about something yeah.