Roblox CEO David Baszucki admits that keeping up with the pace of AI research has become overwhelming—even for someone who once read every paper in his field. Speaking on the "Access" podcast, Baszucki reflected on his recent sabbatical, during which he immersed himself in the latest developments in artificial intelligence. “It was very humbling,” he said. “It is hard to get to a position where you can understand all of those research papers.” Baszucki founded Roblox in 2005, and in its early days, he prided himself on staying deeply engaged with emerging technologies. Back then, he read extensively on topics like physics simulation and rendering, and felt confident in his ability to grasp the cutting-edge ideas shaping the future of gaming. “People were thinking about all of these new ways of doing gaming, and generally I could understand all of it,” he recalled. But that changed with what he describes as the “horizontal wave” of AI research—rapid, broad, and relentless. From transformers and diffusion models to world models and multimodal systems, the volume and complexity of new breakthroughs have surged. “There’s a lot going on,” Baszucki said, acknowledging that the field has evolved far beyond the scope of any single individual’s expertise. The shift reflects a broader transformation in AI. Once primarily an academic pursuit, AI research has become a strategic priority for major tech companies like Meta, Microsoft, and Google. These firms have built large in-house research teams and offer top-tier compensation to attract elite talent. As competition intensifies, some companies are also moving toward more closed research practices. In 2023, Google announced it would reduce the number of AI research papers it publishes, citing a shift toward protecting proprietary knowledge. A Google Brain engineer told Business Insider at the time: “Now it's time to compete and keep knowledge in house.” Despite this trend toward secrecy, OpenAI co-founder Ilya Sutskever recently argued that fundamental research remains the key to unlocking AI’s next breakthroughs. “It's back to the age of research again, just with big computers,” he said, underscoring the enduring importance of innovation over scale alone. For Baszucki, his deep dive into AI research led to a clear insight: the 3D space—central to Roblox’s mission—is still in its infancy. He described AI as a “very physically unnatural space,” trained primarily on human-generated text and images rather than the raw, physical data of the real world. “We're training AI models on this thing we made up,” he said, “rather than on the 3D raw material of the world itself.” This gap, he believes, represents both a challenge and a significant opportunity for the future of immersive digital experiences.