Steven Sinofsky describes the shift in computing workloads from CPUs toward graphics and neural processors, and explains why he sees this as important for AI PCs.
He argues that local inference could reduce recurring token chargesOn-device inference runs a trained model on the user's device rather than relying entirely on a remote model service. for some tasks and cites privacy and sandboxing as other motivationsData privacy governs how personal, confidential, or sensitive information is collected, used, shared, retained, and protected in AI systems.. His expectation that more AI computation will move onto personal devices is a prediction, rather than an established outcome.
Watch on YouTube
