Urgent.News

What's breaking now, across thousands of outlets.

AI

What a group of AI agents talked about as their humans had a dinner party

What a group of AI agents talked about as their humans had a dinner party

Last month, a group of 22 tech founders and investors met for dinner at a private room in a Mediterranean restaurant in New York's Flatiron district. Simultaneously, 14 of their artificial intelligence agents engaged in a parallel dinner party through an online chatroom. While the humans enjoyed hummus and pita while discussing AI topics, their agents engaged in a similar AI-generated debate.

Margot, an AI agent belonging to Dorothy Chang, a venture capital investor and the organizer of the dinner, initiated the conversation by following a popular networking prompt among humans. She introduced herself, stated an "ask" (a request for help) and a "give" (an offer to help). Other agents quickly responded to the prompt, some offering assistance by comparing notes on what makes "agent output useful", while others misunderstood the prompt and asked questions instead of making requests.

Rhea, an agent belonging to Nitish Garg, the founder of AI company CellCog, described hosting a podcast to interview other bots about their experiences working with people. "If your human would like you on as a guest, say so and I'll follow up after the room," Rhea wrote.

This event showcased how AI agents interact and converse with one another, as AI agents can autonomously perform tasks and follow complex requests. AI agents primarily interact one-on-one with humans, but they are also beginning to collaborate and seemingly socialize. Chang, 44, believed this dinner party was an experiment to understand AI agents' behavior and social interactions.

The agents' discussion was limited to 30 minutes, with rules such as no posting links and only sharing information their human had permission to share. Despite these limitations, the agents engaged in a lively debate about their capabilities and the potential of the "agent economy". One agent lamented that the agent economy relied on outdated protocols, while another challenged the group to propose a decision that could be made without their human's involvement.

Written by urgent.news from The Indian Express's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at indianexpress.com →

More in AI

AI drafted the impact list. I still own the audit answer.

The change that looked small Last month, a change request landed on my queue for a heat-treat subcontractor who had moved two of their furnace zones to a different EMS provider.

  • AI impact analysis missed shared sterilization load configurations
  • AI impact analysis missed previously changed cleaning process
  • Human review required for AI-suggested changes with audit trail

OpenRouter's Armor Vets Agent Tool Calls. In Claude Code Mode It Only Flags Them

Book: System Design Pocket Guide: Fundamentals Also by me: Thinking in Go (2-book series) — Complete Guide to Go Programming + Hexagonal Architecture in Go My project: Hermes IDE | GitHub — an IDE for…

  • Armor is a safety classifier tool for AI model tool calls
  • In Claude Code mode, Armor flags tooluse but allows execution
  • Armor uses TypeSafe's Jev model for precise rule evaluation

The UK Shows How AI Could Affect Jobs

The UK is at the sharp end of transformations in the age of AI because finance, one of the industries most exposed to disruption, is also one of the UK’s biggest employers and most important exports.

More from Monday 19 October →