Part 9/12:
Shapiro emphasizes that what Extropic has demonstrated with the TSU is reminiscent of the early days of deep learning—similar to how GPT-2 represented a significant step towards artificial general intelligence (AGI). Despite being in nascent stages, the room-temperature, scalable nature of TSUs makes it a tantalizing glimpse into a future where such devices could become mainstream.
Comparing to Existing Computation Paradigms
Thermodynamic computing sits between analog systems and digital GPUs. While GPU-based simulations approximate noise computationally, TSUs inherently leverage physical noise as a resource. This fundamental difference enables a level of efficiency and parallelism impossible with conventional digital methods.