© IE Online Media Services Pvt Ltd
Tags:
Journalism of Courage
A few moons ago, someone wise said that ChatGPT was just the beginning, and they couldn’t be more correct in their assertion. Stanford University and Google Research recently unveiled their groundbreaking research. Since time immemorial, there have been countless literary works, sci-fi movies, and shows that brought a future world of interactive simulation, precisely a world where humans and humanoid robots co-existed with relative ease.
Well, the new research from Stanford and Google seems to be indicating that the days when humans and robots will be part of a larger simulation may not be far away after all. The revolutionary study, Generative Agents: Interactive Simulacra of Human Behavior, brings forward the concept of generative agents, which are essentially computational software agents. The study is considered to be a turning point in AI as the generative agents are capable of simulating ‘believable human behaviour’.
Based on the study, generative agents are computational software agents that use artificial intelligence to simulate believable human behaviour. These agents draw on generative models to imitate individual and group behaviours. These behaviours are human-like and are based on their ‘identities, environment, and changing experiences’.
As part of the study, researchers from Stanford and Google created an interactive environment that is inspired by the popular simulation video game, The Sims. The sandbox environment was created to demonstrate the capabilities of these AI agents. Researchers later added 25 generative agents or AI agents to the environment. All these agents have different personalities, skills, preferences, and goals. They also have memories of past experiences and can reflect on current situations. Moreover, they are capable of performing various tasks and can react to events.
The generative agents follow the architecture, essentially a large language model (LLM) that consists of components such as observation, reflection, and planning. The architecture can also store records of an agent’s experiences with the help of natural language. Based on the research paper, users were able to interact with the agents via natural language. Besides, the generative agents were also endowed with the ability to strike up conversations which were mostly based on their observations, reflections, and plans. According to the study, the ability to create realistic and consistent behaviours which reflect their personalities, skills, and preferences is the yardstick to measure the believability of these AI agents.
From movies such as Ready Player One or the popular show Westworld, we have often wondered about worlds that are inhabited by virtual yet realistic characters. These characters, although virtual, piqued our interest, kindling in us a need to interact with and understand their everyday lives. Perhaps, this is the fundamental idea of generative agents.
Based on the sandbox experiment, it was learned that generative agents can make breakfasts, commute to work, paint, form relationships, share opinions, initiate conversations, etc. They are eerily similar to humans as they are able to remember, retrieve, and reflect on past and present.
The sandbox environment allowed researchers to study the individual and social behaviours of AI agents. Some of the agents demonstrated human-like preferences and goals. For example, Agent Alice is a painter who loves arts and is striving to improve her skills. She works as a waitress at a diner and has a crush on another AI agent, Bob, who is a writer working in a bookstore. Alice spends her leisure time painting at her home or visiting museums. In order to impress her crush, Alice visits Bob at his work and shows him her paintings.
Another AI agent Eve is a journalist who works for a local daily. She is adventurous and likes to investigate and write stories. Eve is friends with photographer Frank who is her colleague; the duo works together on several projects. Charlie, another agent, is a personal trainer at a gym. He loves working out and enjoys eating healthy foods. Charlie is shown to have a rivalry with Dave, another trainer at the gym.
The above are a handful of the agents from the 25 who inhabit the sandbox environment. Interestingly, each agent has its own personality and stories that users can discover and engage with.
Generative agents are an innovative concept that combines the best of AI and interactive simulation. They have the ability to create realistic simulations of human behaviour that users can explore.
Just like any other technology, generative agents too need thoughtful consideration when it comes to their ethical implications. This is precisely because the whole idea of humans coexisting with virtual beings comes with a host of shortcomings. If we take into account the narrative of the popular show Westworld, what began as benign experimentation soon snowballs into a dystopian world. If the technology, sooner or later, is made available to the public, it has the potential to shake the moral compasses of humans, as the distinction between real and simulation is likely to be blurred to a great extent.
On the contrary, the potential applications of generative agents could also aid in devising new avenues of entertainment, especially for creating immersive virtual worlds. They may also be used for education, particularly in teaching social skills. Moreover, they can aid in research on human behaviour. As of now, generative agents are not perfect and come with a host of limitations, for example, they are prone to errors and inconsistencies owing to the limitations in the language model or simulation environment.
Regardless of the fact that they may raise numerous ethical and social issues, generative agents are believed to be a step towards the future of AI.
