HackerPapers

arXiv:2304.03442cs.HC

Generative Agents: Interactive Simulacra of Human Behavior

Joon Sung Park, Joseph C. O'Brien, Carrie J. Cai, Meredith Ringel Morris, Percy Liang, Michael S. Bernstein

Abstract

Believable proxies of human behavior can empower interactive applications ranging from immersive environments to rehearsal spaces for interpersonal communication to prototyping tools. In this paper, we introduce generative agents--computational software agents that simulate believable human behavior. Generative agents wake up, cook breakfast, and head to work; artists paint, while authors write; they form opinions, notice each other, and initiate conversations; they remember and reflect on days past as they plan the next day. To enable generative agents, we describe an architecture that extends a large language model to store a complete record of the agent's experiences using natural language, synthesize those memories over time into higher-level reflections, and retrieve them dynamically to plan behavior. We instantiate generative agents to populate an interactive sandbox environment inspired by The Sims, where end users can interact with a small town of twenty five agents using natural language. In an evaluation, these generative agents produce believable individual and emergent social behaviors: for example, starting with only a single user-specified notion that one agent wants to throw a Valentine's Day party, the agents autonomously spread invitations to the party over the next two days, make new acquaintances, ask each other out on dates to the party, and coordinate to show up for the party together at the right time. We demonstrate through ablation that the components of our agent architecture--observation, planning, and reflection--each contribute critically to the believability of agent behavior. By fusing large language models with computational, interactive agents, this work introduces architectural and interaction patterns for enabling believable simulations of human behavior.

arXiv comments
Journal reference
DOI
Subjects
cs.HCHuman-Computer Interaction cs.AIArtificial Intelligence cs.LGMachine Learning

On Hacker News

Linked by 25 identifiable commenters in 28 comments across 27 threads, from Apr 2023 to Aug 2026. Found in 4 HN submissions from Apr 2023 to Apr 2024. Open submission discussions.

Comments linking this paper, per year.

Discussed in comments27 threads

  1. Self hosted email continues to steeply decline 1 comment

    • Why do you give up so easily? Human culture and the already low quality bar of writing should be allowed to fall even further because you are annoyed by those that want to protect it? Why should your opinion be worth more than those who would like to retain a semblance of…

      genxyView on HNParent context

  2. Show HN: Context Gateway – Compress agent context before it hits the LLM 1 comment

    • Your comment reminded me of this old simulacra paper (https://arxiv.org/pdf/2304.03442) :) iirc, they compressed the "memory roll" of the agents every once in a while

      ivzakView on HNParent context

  3. Show HN: What if AI agents had Zodiac personalities? 1 comment

    • Minsky's key works on this: "K-lines: A Theory of Memory" (1980) -- Cognitive Science. Names as activation vectors that reactivate entire constellations of knowledge. https://www.sciencedirect.com/science/article/abs/pii/S03640...…

      DonHopkinsView on HNParent context

  4. Total War: Pharaoh's Nvidia Ace AI Advisor Is a VRAM Gobbler 1 comment

    • 6GB of VRAM isn't that much for an LLM. The game itself recommended GPU is a 6GB part. It's kind of surprising that an in game LLM hasn't been done outside of research. It been almost 3 years since "Generative Agents: Interactive Simulacra of Human Behavior"…

      kristianpView on HNParent context

  5. The New AI Consciousness Paper 1 comment
  6. Creating Rimworld Mods to Add LLM PCG and LLM NPCs to Enhance Its RPG Elements 1 comment

    • In my own system in unity, I used LLMs as merely an orchestrator and left Behaviour Trees (https://arxiv.org/abs/1709.00084) as the main core of NPC behaviour. I assume this question is directed at the NPC engine part of the mod, "FelPawns" as of right now. Since my goal is to…

      walterfreedomView on HNParent context

  7. TinyTroupe, a new LLM-powered multiagent persona simulation Python library 1 comment
  8. Project Sid: Many-agent simulations toward AI civilization 2 comments

    • > LLMs are stateless and they do not remember the past (as in they don't have a database), making the training data a non-issue here. That's not what they said. They said that a LLM knows what elections are, which suggests they could have the requisite knowledge to act one out.…

      PhilpaxView on HNParent context

    • > I've always dreamed of full blown sandbox games with extremely advanced NPCs (which the current LLMs can already kinda emulate) The future of gaming is going to get weird fast with all this new tech, and there are a lot of new mechanics emerging that just weren't possible…

      drusepthView on HNParent context

  9. Ask HN: What are you working on (August 2024)? 1 comment

    • I and a friend are working on Elixir Emporium [0], a single-player crafting RPG that uses layers of generative AI to create and inject new game data as you play. It started out as just a prototype asking what happens if you task an LLM with generating crafting recipes for every…

      drusepthView on HNParent context

  10. Dolphin-2_6-Phi-2 1 comment
  11. Lamini LLM Finetuning on AMD ROCm: A Technical Recipe 1 comment

    • > When a new request comes in, a few pointers in HBM are updated in approximately 2.93us, making model switching nearly instant. This enables one server to support more than 10,000 different models. I am a PhD student wanting to work with systems made of multiple LLM powered…

      dhruvdhView on HNParent context

  12. LLM Agent Paper List 1 comment

    • I think about agents as being an entity that you can collaborate with. Among other things, it needs 1/ memory, 2/ ability to create and execute long-term plans. One of the best papers that I have read on this topic is the Generative Agents paper by Park [1] Today's LLMs are kind…

      localhostView on HNParent context

  13. A literary history of fake texts in Apple's marketing materials 1 comment

    • The article's funny, and kind of surreal read considering a dimension fading away while all the characters become cartoons and the landscape becomes vague. Have similar thoughts occasionally. However, also creepy. Not sure if its the way humans have always written, that ChatAI…

      araesView on HNParent context

  14. Show HN: WhatsApp-Llama: A clone of yourself from your WhatsApp conversations 1 comment

    • Thank you, Ill check it out! Yeah, this can be extended to create a "simulation game" of us and our friends. This paper (Interactive Simulacra of Human Behaviour https://arxiv.org/abs/2304.03442 ) has a setup on how we could create a Sims game with us as the characters

      advaith08View on HNParent context

  15. Ask HN: Which recent research paper blew your mind? 1 comment

    • Generative Agents: Interactive Simulacra of Human Behavior[1]. Make sure to check out the recorded demo! [1] https://arxiv.org/abs/2304.03442

      dan-gView on HNParent context

  16. AI: Startup vs Incumbent Value (2022) 1 comment

    • > It just never got widespread adoption because it's just not interesting. True. There have been NPC systems where the NPCs had motivations and a life of their own, even when no one was around. Those haven't helped gameplay much. The current problem is that LLMs don't know…

      AnimatsView on HNParent context

  17. Roleplaying with LLMs 1 comment
  18. The future of AI according to thousands of forecasters 1 comment

    • >GPT cannot function in any environment on its own. Except in response to a direct instruction it cannot plan, it cannot take action, it cannot learn, it cannot adjust to changing circumstance. Sure it can. It's not default behavior sure but it's fairly trivial to set up, just…

      famouswafflesView on HNParent context

  19. ChatGPT is really good at roleplaying 1 comment

    • This is naive but has at least one small improvement. I'll call this "level 1.5" roleplaying. - Level 0 is just using chat.openai.com. - Level 1 is just putting character description in the system prompt - Level 2 is doing third person prompts, like "you are writing dialog for…

      ianbickingView on HNParent context

  20. We aren't close to creating a rapidly self-improving AI 1 comment

    • What you describe sounds like experiments with "generative agents". Check out https://arxiv.org/abs/2304.03442

      arc619View on HNParent context

  21. Autonomous Agents and Agent Simulations 1 comment

    • Research on Agent architectures is evolving real fast. I read Park et al.'s paper on Generative Agents a couple of weeks back which seems to have the potential to put LangChain's Memory on steroids: https://arxiv.org/abs/2304.03442 "To enable generative agents, we describe an…

      akhayamView on HNParent context

  22. Emergent abilities of large language models 1 comment

    • Yeah i honestly don't think you realize where we're at already. It's not even about gpt-4 honestly. >and in the case of GPT-4, it's pretty evident that it does not have such consciousness. Hard disagree but i guess that's what people might think with all the "as a large language…

      famouswafflesView on HNParent context

  23. Surprising things happen when you put 25 AI agents together in an RPG town 1 comment
  24. What if someone mixed The Sims with ChatGPT bots? It would look like this 1 comment

    • Original paper: https://arxiv.org/abs/2304.03442 Heroku demo: https://reverie.herokuapp.com/arXiv_Demo/#

      epupsView on HNParent context

  25. Stanford's Groundbreaking AI Study Simulates Authentic Human Behavior 1 comment
  26. Generative Agents: Interactive Simulacra of Human Behavior 1 comment
  27. With LLMs, Researchers create Generative agents that interact with each other 1 comment

Submitted to Hacker News

  • Show HN: AI-town, run your own custom AI world SIM with JavaScript429 points115 comments

    Hi HN community! We want to share AI-town, a deployable starter kit for building and customizing your own version of AI simulation - a virtual town where AI characters live, chat and socialize. Inspired by great work from the Stanford Generative Agent paper (https://arxiv.org/abs/2304.03442). A few features: - Includes a convex.dev backed server-side game engine that handles global state - Multiplayer ready. Deployment ready - 100% Typescript - Easily customizable. You can fork it, change…

  • Generative Agents: Interactive Simulacra of Human Behavior391 points252 comments
  • Show HN: LLM-powered NPCs running on your hardware24 points4 comments

    Hey HN, My team is open-sourcing the inference stack and fined-tuned models we use to create LLM-powered NPCs: https://github.com/GigaxGames/gigax The generative agents paper [1] pioneered the idea of prompting LLMs to create autonomous NPCs. But existing implementations require multiple calls to an LLM to make the agent plan its day, chat with people, and interact with its environment [2]. Our approach allows NPCs to be stepped at runtime with a single pass on consumer-grade hardware, with…

  • Generative Agents: Interactive Simulacra of Human Behavior13 points2 comments

Connected through shared HN threads, not inferred similarity or a recommendation.