On 8/5/07, Randall Shimizu <[EMAIL PROTECTED]> wrote: > Well just imagine how much processing power it require to simultaneously > interpret 100,000 phone calls at once. > > ----- Original Message ---- > From: Tracy R Reed <[EMAIL PROTECTED]> > To: Main Discussion List for KPLUG <[email protected]> > Sent: Sunday, August 5, 2007 2:13:15 AM > Subject: Re: Google builds own phone | The Register > > Randall Shimizu wrote: > > If Google were to do this, it would require massive server farms or Google > > might be forced to switch to RISC based systems since this a very numeric > > intensive process. > > A massive server farm...hmmm...let me see...where might google get a > massive server farm... > > -- > Tracy R Reed
For all of the hype serious speaker independent voice recognition just ain't there yet. Forget the problem of needing massive horsepower. But if you want a lot of horsepower why not look at the local phones themselves. Say Google decides to give away phones subject to the condition that Google gets to run apps on the phones when you are not using them. I can envision a lot of possibilities with that model. Basically Google starts building another massively distributed mobile computer farm. Now this is not without its problems; battery power comes to mind immediately. But the properties of such a distributed system must map well onto some classes of application, like say apps that need locally oriented data. Use the local phones as distributed caches for that data and only go out to the big "server in the distant cloud" if you miss hitting what you need in the local cloud. Or Google could say tie the phones tightly to local servers, e.g. in every wireless access point (Starbucks, etc.) Lots of possible ways to get innovative with both architectures and business models here. Serious engineering study of the properties of these sorts of networks would be a _lot_ of fun. BobLQ -- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
