Skip to content
Last Dictionary

generative agent

nounAlso: generative agents, Smallville, generative agent simulation

In plain wordsA character in a computer simulation, run by a language model, that keeps a diary-like memory, draws lessons from it and schedules its days, so that a group of them can act like a plausible small community.

  1. A software character driven by a large language modellarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code. and given memory, reflection and planning so that its conduct stays humanlike over long stretches. In the design of Joon Sung Park and colleagues (2023), every observation is logged in plain language in a 'memory stream'; a retrieval step ranks memories by recency, importance and relevance to choose what enters the prompt; the agent periodically condenses memories into higher-level insights; and it drafts and revises plans at several time scales. Twenty-five such agents in a Sims-like sandbox town kept daily routines and built relationships, and one seeded idea — a resident's plan for a Valentine's Day party — spread by word of mouth until guests arrived together. Generative agents are used for social-science simulation, games, user research and policy modeling; critics caution that simulated people may reproduce stereotypes and should not be mistaken for real survey respondents.

    We piloted the questionnaire on generative agents first, then checked their answers against real respondents.

Origin

Introduced by Joon Sung Park, Joseph C. O'Brien, Carrie J. Cai, Meredith Ringel Morris, Percy Liang and Michael S. Bernstein (Stanford University and Google) in 'Generative Agents: Interactive Simulacra of Human Behavior', posted on 7 April 2023 and presented at the ACM UIST conference in October 2023.

Usage note

The memory-stream-with-reflection design influenced later agent memoryagent memory, n.: The ways an AI assistant or agent keeps information — your preferences, past conversations, lessons from earlier tasks — and brings it back later, beyond what fits in a single conversation. systems. Not to be confused with a generative modelgenerative model, n.: A model that learns what a kind of data looks like — faces, sentences, songs — well enough to produce new examples that resemble the originals without copying them. in general.

Related

See also
  • agent memoryagent memory, n.: The ways an AI assistant or agent keeps information — your preferences, past conversations, lessons from earlier tasks — and brings it back later, beyond what fits in a single conversation.
  • multi-agent systemmulti-agent system, n.: A setup in which several AI agents, each with its own role or point of view, work together — or compete — on a problem, passing messages and results among themselves.
  • AI agentAI agent, n.: An AI system that does not just answer but acts: given a goal, it decides on steps, uses tools such as a browser or code editor, checks the results and keeps going until the job is done.
  • long-term memorylong-term memory, n.: Memory that lasts. For an AI agent, it means information saved outside the conversation — facts, past events, learned procedures — that can be looked up and used again days or months later.
  • simulator theorysimulator theory, n.: A way of thinking about AI language models as simulators that can play any character found in their training text, rather than as a single agent with its own beliefs and goals.

Sources

  1. Park, J. S., O'Brien, J. C., Cai, C. J., Morris, M. R., Liang, P. & Bernstein, M. S. (2023). Generative Agents: Interactive Simulacra of Human Behavior. arXiv:2304.03442 (UIST 2023). arxiv.org/abs/2304.03442

Mentioned in

multi-agent systemmulti-agent system, n.: A setup in which several AI agents, each with its own role or point of view, work together — or compete — on a problem, passing messages and results among themselves., sandboxsandbox, n.: A sealed-off computer space where an AI agent can run programs and change files without being able to harm the real system, reach private data or contact the outside world unless allowed.