My Photo

November 2015

Sun Mon Tue Wed Thu Fri Sat
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24 25 26 27 28
29 30          


« Do Our Schools Really Suck? (Working Title) | Main | 'The Next Grand Minimum' Launches »

19 August 2011


Douglas Keachie

Big Improvement over previous post.

Douglas Keachie

Hey, hey, hey, here comes Colossus!

Ryan Mount

Well yeah, but can it play chess and beat the pants off of Celebrity Jeopardy contestants? Probably.

"The 9000 series is the most reliable computer ever made. No 9000 computer has ever made a mistake or distorted information. We are all, by any practical definition of the words, foolproof and incapable of error." ~HAL

George Rebane

Good point RyanM, but we have to remember that HAL was represented as a human-programmed AI, and even then Arthur Clarke did not assert that HAL committed an error. If anything, it was HAL's human handlers who seemed to have overlooked providing the machine with a proper utility function within which it should make its decisions. The one HAL was given, or cooked up for itself, clearly demonstrated that it acted reasonably.

Ryan Mount

Good point about HAL, BTW. I was thinking more about hard drive failures, etc. But given enough redundancy and fail over, such a CPU might be able predict and adapt to errors and crashes. We have that now to some extent (disclosure: I work for IBM), but IMO it's largely a marketing exercise with PowerPoint Slides with a lot (Data Centers worth) of hardware thrown at it. And it's very expensive.

In all seriousness though, the application for such a processor are endless. It could detect human error by analyzing patterns of behavior, rather than just crunching numbers for analysis by a human. Imagine a future version of Microsoft Word that could coach your writing style, much like a teacher might by analyzing large chunks of prose.

LOL. It could put economists out of work. And bloggers, I guess. ;-)

George Rebane

RyanM - here's a 'secret agent request' - can you find out or point me (and interested readers) to an IBM Research URL that describes the structures used in this CCC. I'm particularly interested if IBM is supporting or adapting something from/with Numenta's HTM technology that was first described Jeff Hawkins.

As a look-ahead, my next question would be how such brain-similar structures would simplify the implementation of Bayes nets (since actual critter brains do that very well).

Ryan Mount

Most of the Research comes out of the Alamaden Research Center down by San Jose:

I do know that they have an entire team dedicated to something called Cognitive Computing:

I find it interesting that they are using the human brain as a template for this project. So rather than creating something from scratch, they're attempting to some extent mimic the "functionality" of the human brain.

It's a very interesting epistemological question/problem and it's certainly one that has daunted many post-modern thinkers: How do we know what we know? what are the limits of our knowledge? Is there such a thing? Perhaps through the aid of technology (enter: Arthur C. Clarke), we might be able to surpass the capabilities of the human brain.

George Rebane

Exactly so RyanM, and that is the point of the Singularity Signposts section on RR. And thank you for the input and links.

The comments to this entry are closed.

Blog powered by Typepad