@cstross @skjeggtroll Moore’s Law is dead for now. I did a study a few years ago to look at what is happening and will happen in microprocessors. Short story is that traditional processor architecture is hitting end of life. Feature sizes are so small now that quantum effects are a significant factor. High speed and small size also means we are up against a thermal barrier as well. Clever approaches with System-On-a-Chip ( #SOA), 3D stacking, maybe Processor-In-Memory ( #PIM) and distributed multiprocessing will squeeze out more progress for maybe a decade. After that comes the next computing revolution — a shift to non-Von Neumann #computing. #Quantum has the spotlight because that’s the really big win, but there are other approaches that are likely to be commercially viable before quantum is mature. I’m optimistic about the tech, less so about the rate of adoption and change that will be required, especially if the most talented early- career computer scientists and engineers keep chasing the associative/statistical methods that include LLMs.

@cstross @skjeggtroll Moore’s Law is dead for now. I did a study a few years ago to look at what is happening and will happen in microprocessors. Short story is that traditional processor architecture is hitting end of life. Feature sizes are so small now that quantum effects are a significant factor. High speed and small size also means we are up against a thermal barrier as well. Clever approaches with System-On-a-Chip ( #SOA), 3D stacking, maybe Processor-In-Memory ( #PIM) and distributed multiprocessing will squeeze out more progress for maybe a decade. After that comes the next computing revolution — a shift to non-Von Neumann #computing. #Quantum has the spotlight because that’s the really big win, but there are other approaches that are likely to be commercially viable before quantum is mature. I’m optimistic about the tech, less so about the rate of adoption and change that will be required, especially if the most talented early- career computer scientists and engineers keep chasing the associative/statistical methods that include LLMs.