Barumonk
Artificial Intelligence - Part 3; "Can you hear me now?"
by
, Sat, 10-29-2011 at 12:52 PM (33238 Views)
This is kind of messy, I think. It's really a continuation of what I was writing in response to Kraco.
Hollywood Undead aside, mobile technology is what will eventually allow you to have your own personal robotic AI in your house or at your work. Yes, that smartphone in your pocket will be the same stuff that allows your robotic AI to greet the Jahovah's Witness and engage in a humorous conversation about god, but we will save the morals of this for the next blog entry.
Why will the technology used in your smartphone by used by AI? Let's see, it's small and lightweight, it has low power consumption unless you're running the MP3 player while browsing the web, it has several different types of common ports to connect to different equipment around the world. It comes with it's own built in telephony system to allow it to place emergency calls or even get really creepy and let you talk to someone through your robot, it's getting faster all the time due to a short development cycle, and it's already in such high demand that they are pretty cheap. Yeah, they are cheap, if you actually look up the cost of a smartphone board with everything including a screen they average about $50. Try to keep your jaw off the ground when you think about how much you're paying for your phone.
Most power consumption on a phone comes from extra features. For instance watching a 1080p movie causes it to turn on a special part of the chipset which handles 1080p movies specifically and drains power like no tomorrow. The actual problem comes in with a single issue: Most mobile processors are based on the ARM specifications, which are 16 bit - not 32 or 64 like most current PCs. This limits the amount of memory available by a large amount. This is why you generally only see 4GB max on a mobile device. Intel is fixing this to some extent with x86 mobile boards, but it's not catching on since most of the world is commited to ARM... or more specifically commited to Google and Qualcomm. Android (the mobile OS) gained popularity so fast because Qualcomm jumped on immediately and started designing chipsets specifically for Android, and these chipsets use ARMv7 for the most part. Other companies loved this because there was no fee to use Android and they didn't have to develop a new chipset for it. They just buy the chipset from Qualcomm and then slap a case on it and then make sure their own company logo is all over the damn thing.
Anyway, the lack of memory is going to be the biggest stepping stone for awhile, at least until the mobile world can find a real use for more memory. Regardless it's still the most promising area of AI that really has nothing to do with AI. Another point is that Microkernels, which are traditionally more stable and secure operating systems, are commonly used by mobile devices. Microkernel development was honestly a dead stick until smartphones blew up. More recently there has been experimentation at MIT with an Exokernel, which is a super lightweight implementation of a Microkernel. This is the technology that your future android/robotic AI will be running on.
So, when will all this be ready for AI? Probably within ten years, maybe a bit more but not past fifteen. This doesn't mean we will have a strong AI within ten to fifteen years, but the potential will be there for much more than what we have right now.
- By part 5 I intend to start getting more specific into AI concepts and how to make them. -