Contrary to common perception shaped by their use in language, Transformers are not inherently sequential. Their core architecture operates on sets of tokens, with sequence information only injected via positional embeddings. This makes them powerful for non-sequential data like 3D objects or other unordered collections.
This idea posits that language is a lossy, discrete abstraction of reality. In contrast, pixels (visual input) are a more fundamental representation. We perceive language physically—as pixels on a page or sound waves—and tokenizing it discards rich information like font, layout, and visual context.
The plateauing performance-per-watt of GPUs suggests that simply scaling current matrix multiplication-heavy architectures is unsustainable. This hardware limitation may necessitate research into new computational primitives and neural network designs built for large-scale distributed systems, not single devices.
World Labs argues that AI focused on language misses the fundamental "spatial intelligence" humans use to interact with the 3D world. This capability, which evolved over hundreds of millions of years, is crucial for true understanding and cannot be fully captured by 1D text, a lossy representation of physical reality.
Fei-Fei Li expresses concern that the influx of commercial capital into AI isn't just creating pressure, but an "imbalanced resourcing" of academia. This starves universities of the compute and talent needed to pursue open, foundational science, potentially stifling the next wave of innovation that commercial labs build upon.
While a world model can generate a physically plausible arch, it doesn't understand the underlying physics of force distribution. This gap between pattern matching and causal reasoning is a fundamental split between AI and human intelligence, making current models unsuitable for mission-critical applications like architecture.
With industry dominating large-scale compute, academia's function is no longer to train the biggest models. Instead, its value lies in pursuing unconventional, high-risk research in areas like new algorithms, architectures, and theoretical underpinnings that commercial labs, focused on scaling, might overlook.
Unlike video models that generate frame-by-frame, Marble natively outputs Gaussian splats—tiny, semi-transparent particles. This data structure enables real-time rendering, interactive editing, and precise camera control on client devices like mobile phones, a fundamental architectural advantage for interactive 3D experiences.