AI agents are currently on everyone's lips.
They can independently decide on the execution of a system's functions, start programs, transmit data, etc.
This raised the question whether agentic behavior can also be used in interactive real-time programs.
For an interactive 3D world in which a user can ask questions to an AI avatar, there are several possibilities for using agentic behavior.
In the interactive demo “Babelsberg – the cradle of filmmaking,” the AI uses communication with the user to classify the content of the conversation. The program contains media files (videos, images, 3D models) on various topics,
which are independently accessed by the AI when it concludes that, for example, the user's question matches a stored topic.
The interactive learning experience about the film studio in Potsdam Babelsberg contains a wealth of background information
on the history of the UFA and DEFA studios.
Each user question is first compared with the stored data, and the relevant information is passed on to the AI as source material for a response.
In a further step, the AI is asked to categorize the questions and compares the stored media with the current conversation content. If the AI finds a match, the media is integrated into our real-time 3D
scene, i.e., videos are played, or 3D models are inserted, which can then be explored by the user.
If the user asks the AI in our app about the “greatest film from UFA Studios,” the AI can decide whether or not to display the media
on the topic of “Metropolis.”
This creates an additional dynamic that makes the learning experience more exciting and enriches the content.
In a real-time 3D game, however, there must be strong control through AI programming (prompting) so as not to destroy the experience and plunge the 3D world into chaos.
In earlier learning programs, content was always used in the same way in a segment and executed schematically. It was hardly possible to modify the program based on a user's question. Here, all possible variations of questions and answers had to be specified in advance.
In the end, the question remains: “Is agentic AI worthwhile in a real-time 3D environment?”
I think so. The results of the Babelsberg demo are promising. The experience gained from this project will help to produce even better
applications that create an individual experience for the user while conveying important information in a controlled manner.
These technologies can be used profitably for training situations, educational environments, or entertainment.
Demo Video : https://www.youtube.com/watch?v=ZyKWrzt3nGM
Christof Wurzer
Top comments (0)