The artificial intelligence industry faces a critical shortage of high-quality data needed to train more powerful models. A new startup, Moonlake AI, aims to solve this problem by creating data from interactive 3D worlds. Rashi Shrivastava reports for Forbes. The company was cofounded by former Stanford PhD students Sharon Lee and Fan-Yun Sun. They are developing AI software that allows users to easily create virtual simulation environments.
The cofounders believe people will use the tool to build worlds for gaming, animation, and education. According to Sun, this activity will organically generate vast amounts of complex data. This data could then be used to train advanced reasoning models, which are AIs designed to solve problems through multiple steps. Lee stated that large-scale interactive worlds represent “the next paradigm that allows you to scale the data infinitely.” The technology can also help AI researchers in fields like robotics create digital simulations to test and verify tasks.
Moonlake AI recently came out of stealth with $28 million in seed funding from investors including AIX Ventures and Nvidia Ventures. Other companies, such as World Labs and Runway, are also working on generating interactive 3D environments.