Roblox, the popular online gaming platform, has introduced a new tool that uses artificial intelligence (AI) to create immersive 3D environments for its in-game worlds. This tool, known as “tokenizing,” treats the 3D blocks that make up these worlds as units that can be assigned numerical values based on their likelihood of coming next in a sequence. This approach is similar to how a large language model, such as GPT-4, predicts the next token in a sequence of words.
To understand how this works, imagine inputting the phrase “The capital of France is …” into a language model. The model would likely predict that the next token is “Paris.” Roblox’s system applies a similar concept to 3D blocks, creating the environment by generating the most probable next block based on the sequence of blocks.
However, developing this tool presented several challenges. One of the main difficulties is the limited availability of high-quality 3D data compared to text data. To overcome this, Roblox had to rely on user-generated data from creators as well as external data sets to train its models.
Anupam Singh, Vice President of AI and Growth Engineering at Roblox, explains, “Finding high-quality 3D information is difficult. Even with comprehensive data sets, predicting the next cube requires considering its three dimensions: X, Y, and Z.”
The scarcity of 3D data can result in unexpected and implausible scenarios, such as a randomly placed tree in the middle of a racetrack. To address this issue, Roblox utilizes a second AI model trained on more abundant 2D data obtained from open-source and licensed data sets. This model, referred to as the 2D AI, verifies the logical consistency of the generated 3D environment by converting it into a 2D representation. If the resulting images lack coherence, the 3D AI generates new blocks until the 2D AI approves the design.
It is important to note that despite the advancements in AI technology, human involvement remains crucial in crafting enjoyable game environments for Roblox’s millions of players. Chris Totten, an associate professor in the animation game design program at Kent State University, emphasizes the significance of human guidance, stating, “A lot of level generators will produce something that’s plain and flat. You need a human guiding hand.” This notion applies to players utilizing AI-generated content in their creations, as well as game designers who leverage the tool.
Moreover, the introduction of this tool raises questions about what constitutes good player-responsive level design. As AI becomes more integrated into the creative process, it becomes essential to evaluate the balance between automated generation and human input. Roblox’s new tool offers exciting possibilities for level design, but it also prompts a broader discussion about the role of human creativity and oversight in game development.
In conclusion, Roblox’s innovative tool that employs AI for generating 3D environments marks a significant step forward in game design. By tokenizing 3D blocks and using a combination of 3D and 2D AI models, Roblox aims to create immersive and consistent gaming experiences. Nevertheless, the indispensable role of human involvement and artistic guidance in crafting compelling game environments cannot be overlooked. As AI continues to evolve, this tool sparks discussions about the future of game design and the dynamic relationship between technology and human creativity.
Source link
