> The thing about "agentic" AIs is that nobody is thinking of the consequences downstream here. Why would your guests turn up, since they'll be getting loads of random spam? How does the bowling alley handle loads of automated requests? Who do you complain to if the cake is wrong?

>> Fascinating @benthompson vision of the future of the Internet-i.e., "The Agentic Web"

>> Scenario: Before I go to sleep, I tell ChatGPT "Plan my 5-year-old's birthday next Saturday, budget $500. When you've made the reservation, email these 20 ppl a printable invitation to attend. Also my wife wants to go to England in mid-July. Find 5 plausible flights for the family and make several distinct itineraries. Finally, pls edit this work memo."

>> When I go to sleep, the Al agent negotiates slots with two bowling alleys, buys a cake, emails printable invites, plans the trip, copy-edits, etc.

> In the real actual world we already live in, restaurants are already reverting to only taking reservations by phone, or charging huge reservation fees, to stop automated booking. This is with basic bots!

So which is it? Are LLM/Agentic AIs ruining things or just an extension of what has already happened without any LLMs involved?

"Hey look at this horrible thing that LLMs can do! I mean, people have been automating this for years successfully but now LLMs can do it as well! The sky is falling!" /s

I don't think you should use AI for something like this, most "Plan a X trip/event" prompts I find hilarious and would never trust (with today's tech at least, and tomorrow's at a minimum) but this doesn't seem like a very strong argument.

It’s costly signaling, we value what we perceive has cost. Driving the perceived cost to zero will plummet the value.