It's clear that AI can be the catalyst for creating a lot of customer-facing value. (Though I still contend that the current incarnation of large language models ain't it yet.) Requiring a powerful (and power-hungry) GPU in all devices is non-tenable, so having a dedicated NPU is a pretty sleek way forward (that, to be fair, Apple spearheaded).
I'm very curious to see what APIs will be open to developers, because I think "next billion-dollar startup"-style value-generation won't be found in the generative aspect of these models, but rather the promise of automation, synthesis, summarization, agent invocation, and so on.
I'm very curious to see what APIs will be open to developers, because I think "next billion-dollar startup"-style value-generation won't be found in the generative aspect of these models, but rather the promise of automation, synthesis, summarization, agent invocation, and so on.