Mike Becky

The Early Days of AI Computing ➝

Stephen Hackett, on Apple’s focus on Siri’s existing feature-set:

I understand Apple wanting to make sure that Siri’s core functionality of controlling your iOS device keeps getting better. That stuff should be bulletproof, but we’re five years into Siri’s life. The company should be moving past these features and making Siri smarter about the world around us.

Here’s the problem with the statement above: it seems to be under the assumption that voice-based AI systems are no longer in their early stages — that five years is long enough to have the core functionality, essentially, finished. But I just don’t think that’s true. This is a complicated feature that may take far longer to perfect than any other computer science problem we’ve seen before. We’re still, very much, in the early days.