Intel CEO Brian is a man living on the edge

Certainly...

By 2020, every stone will generate GigaBytes/hour of data, especially if wrung for blood.

(Also, Intel Architecture ... will it be with us for ever?)

Has Mr Krzanich ever heard of the term 'Overfitting'?

Maybe him and the rest of the goons interested in harvesting all the data should learn about it, it tends to royally screw up otherwise good predictive analysis. Turns out there is defintely such a thing as too much data, and these companies already have too much data, that is why they can't get anything right..

Intel is going to have to watch their back

For now, Intel's processors and other chips are general enough in nature and quick enough to be usable in what I call "intelligence engineering and analysis." If you look at how I file and track things, "artificial intelligence", "machine learning" (not quite the same thing) and "big data" are about taking data, tossing it into models and algorithms, to derive or infer information. I got my start back in the 1970's when "big data" involved fork-lifts and palettes. All that has changed since the beginning is the density and volume, oh, and the amount of skull-sweat involved. Well and good.

Intel's problem is the current state-of-the-art in the above "disciplines" is mostly tossing darts at the dart board and trying, desperately, to find what works. As a few others in the field have observed, how & why, not so much. That's one reason that the field more resembles an art rather that a science. So, the engineering is very hit and miss. This if fine with general purpose hardware if speed and accuracy aren't paramount. Therein lies the rub. When you really, really need that speed and accuracy you have to look at dedicated designs. Thus we see where nVidia, IBM, Google, and even Microsoft are designing custom hardware.

It's going to be an interesting balancing game for Intel. The one thing they've really got an in with is custom designs using their FPGU's and they certainly don't have a total lock on that look at recent announcements of advances in that section of hardware engineering. Even Microsoft has figured out that. Usually they're further behind the curve.

I mostly use Intel here, I've always had respect for their CPU's even if a bit pricey. However, when it comes to doing intelligence engineering, I bring bigger, more specialized hardware that's, unfortunately, pricier by far. On a performance/price basis, actually cheaper.