3 ms·

(1) Right now they do but it's only a matter of time before Samsung or HTC partner up with Prada, Rolex, Omega or any other major brand and come out with a line

by m_ke 12y ago

(1) Right now they do but it's only a matter of time before Samsung or HTC partner up with Prada, Rolex, Omega or any other major brand and come out with a line of watches that appeal. The point is that we might get one or two apple watches a year if we're lucky and in the same time frame we'll be getting around 30-50 different android wear devices. Android wear will be available in a ton of different shapes and sizes, and there's no way that the top watch brands won't want to get in on the action. (2) Data is everything in machine learning and google has the whole web. Google knows me better than my mother, girlfriend and brother combined. Android, youtube, search, gmail, drive, maps, calendar, hangouts and shopping express are enough to tell them who I talk to, where I live, what I eat, where I'm going and what I read. A couple of percentage points is huge when your accuracy is above 90%. Apple doesn't have anyone as good as Geoff Hinton, Peter Norvig, nor Jeff Dean. There's no way they'll be able compete on AI with google.

achompas 12y ago
(1) Google has partnered with Diane Von Furstenberg on Glass and...there isn't a whole lot to show for it. Glass has ~zero mindshare among fashion-forward people. I cannot stress the importance of design as a core competency. (2) You're the second person to respond with "data is everything" in machine learning, but it's more complicated than that. While Google has Hinton working on deep nets and Kevin Murphy on knowledge representations, this cutting-edge work represents something closer to MS Research or Bell Labs. These models take years to affect production. My experience is that "a couple of percentage points" above 90% actually matters little--the marginal cost of obtaining those points is enormous (many many hidden layers in your convolutional net really slows it down) with little real-world benefit (a user can't tell the difference between 90 and 95%--they'll just think of the product as "really really good"). I believe Apple can approach the point where their NLP/speech tech is "good enough" relative to Google.

by m_ke 12y ago

achompas 12y ago

(1) Google has partnered with Diane Von Furstenberg on Glass and...there isn't a whole lot to show for it. Glass has ~zero mindshare among fashion-forward people. I cannot stress the importance of design as a core competency. (2) You're the second person to respond with "data is everything" in machine learning, but it's more complicated than that. While Google has Hinton working on deep nets and Kevin Murphy on knowledge representations, this cutting-edge work represents something closer to MS Research or Bell Labs. These models take years to affect production. My experience is that "a couple of percentage points" above 90% actually matters little--the marginal cost of obtaining those points is enormous (many many hidden layers in your convolutional net really slows it down) with little real-world benefit (a user can't tell the difference between 90 and 95%--they'll just think of the product as "really really good"). I believe Apple can approach the point where their NLP/speech tech is "good enough" relative to Google.