The progression from early neural networks to today's massive models is fundamentally driven by the exponential increase in available computational power, from the initial move to GPUs to today's million-fold increases in training capacity on a single model.
A common misconception is that Transformers are sequential models like RNNs. Fundamentally, they are permutation-equivariant and operate on sets of tokens. Sequence information is artificially injected via positional embeddings, making the architecture inherently flexible for non-linear data like 3D scenes or graphs.
Fei-Fei Li's lab believed they were the first to combine ConvNets and LSTMs for image captioning, only to discover through a journalist that a team at Google had developed the same breakthrough concurrently. This highlights the phenomenon of parallel innovation in scientific research.
Human intelligence is multifaceted. While LLMs excel at linguistic intelligence, they lack spatial intelligence—the ability to understand, reason, and interact within a 3D world. This capability, crucial for tasks from robotics to scientific discovery, is the focus for the next wave of AI models.
With industry dominating large-scale model training, academic labs can no longer compete on compute. Their new strategic advantage lies in pursuing unconventional, high-risk ideas, new algorithms, and theoretical underpinnings that large commercial labs might overlook.
The performance gains from Nvidia's Hopper to Blackwell GPUs come from increased size and power, not efficiency. This signals a potential scaling limit, creating an opportunity for radically new hardware primitives and neural network architectures beyond today's matrix-multiplication-centric models.
Vision, a product of 540 million years of evolution, is a highly complex process. However, because it's an innate, effortless ability for humans, we undervalue its difficulty compared to language, which requires conscious effort to learn. This bias impacts how we approach building AI systems.
A Harvard study showed LLMs can predict planetary orbits (pattern fitting) but generate nonsensical force vectors when probed. This reveals a critical gap: current models mimic data patterns but don't develop a true, generalizable understanding of underlying physical laws, separating them from human intelligence.