Do you also believe that computers will program themselves some day? There are some pretty fundamental problems with that idea.
The thing that most laypeople don’t understand about AI is that it’s not an artificial brain capable of solving any and all problems we might throw at it. We have this image of computers surpassing human intelligence, taking over the world and turning us into a bunch of energy cells. And who knows, that might still happen. But so far we don’t even know what intelligence is! And compared to that dream, the today’s AI is just a giant toolbox for manipulating data. It’s still a human’s job to know how to use the toolbox and how to make sense of the data coming in and the data getting out. Yes, you can create a very sophisticated algorithm for solving a very specific problem way better than any human. Like flying an airplane, driving a car or trading stocks. BUT! 1) Some human has come up with that algorithm, not a machine. And when an algorithm runs across an unexpected set of inputs, it stumbles - very often quite spectacularly. 2) Anything less specific, especially when interacting with humans is involved, and we still have a looooong way to go. Probably decades before we come up with anything useful.
Meanwhile, the complexity of the Internet of Things in a single home is going to increase exponentially over the next five years. That dream of an “intelligent toaster” is finally becoming a reality. Washing machines, fridges, stoves, microwaves, furnaces, ACs, lights, all sorts of sensors and so on and so forth - we are going to have THOUSANDS if not millions of connected things in our homes (some of us are almost there already). People will realize that “central intelligence” approach to Home Automation is not going to work in the long run and a lot of the intelligence will start moving to the edges - into the individual devices. We’ve been down this path MULTIPLE times already! Perhaps the MQTT-S might even see some traction! And that’s why I’m betting my money on ZigBee and not Z-Wave. (I’m hoping that Samsung acquiring ST means more ZigBee appliances and devices!) And why I also think that WiFi and BLE-connected device manufacturers are missing the point. Although, BLE is really cool, so it’s going to find its niche.
Bottom line, the “central intelligence” route is going to lose in the long run. Obviously, any cloud needs orchestration and IoT is not an exception. But “orchestration” is not “central intelligence”. Think bad manager vs. good manager. Bad manager tends to think of herself as the smartest person in the room, micromanages and gets in the way of everybody else. Good manager hires people smarter than herself, delegates decision making to them entirely, removes any obstacles in their way and makes sure they talk to each other regularly and efficiently. This is what ST must be focusing on in the long run.
And Bill gates starred that we would only need 64k of ram. Things and technologies change. While you are right that we are not ready yet, I do believe that there are many people working toward this goal
Ants are not intelligent by themselves but working together than can move mountains
Unfortunately your thinking is what is making it so hard to develop an AI. There is no need to develop code for all instances. Most challenges are related to navigation in human environments, followed by the actions necessary to perform the required actions. Not saying that coming up with a basic AI is easy, it isn’t. But, we have come a long way and as computing power improves and new technologies become available we will build or dreams.
Meh… all cool kids are playing BT Smart these days.
WiFi + BTS cover all the bases: WiFi for long range, high bandwidth and BLE for short range, low power. There’s no room for Zigbee in this ecosystem. It’s just plain old survival of the fittest. I spent a lot of my professional time working with Zibgee and I used to love this technology, but the truth is that they totally blew it in consumer space. There still may be a hope for Zigbee in lighting applications if enough OEMs get behind LightLink, but all other profiles, particularly SE and HA are as good as dead.
I am sorry for making it so hard for you AI guys. Next time I’ll be thinking of unicorns and rainbows; hopefully that’ll help!
The point I was successfully failing to make is this: AI is not Intelligence. We don’t know what Intelligence is (though we have a bunch of contradicting theories).
I guess there’s only one way to find out who’s right and who’s wrong (no, it’s not Google). We shall wait and see.
Though, like I said, the specific protocols are not important. The industry will converge on one or two complementary protocols (they might as well be WiFi and BLE, I won’t cry too much if ZigBee disappears entirely).
@sudarkoff ohhhhhhh., now I understand. Why didn’t say so in the first place? (SMILE), I do agree with you, but remember what AI stands for Artificial Intelligence, Meaning (not true intelligence) simulated. I thought you were saying that we would never have a working AI, when I have seen fairly decent AI’s at my place of business, and with Health Care, Anthem is using Watson (IBM) to go through all cancer records and work out best plan of attacks given the status of the patient and type of cancer. It is pretty impressive and of course most of us who were interested saw it on Jeopardy.
No it is not perfect, and it runs some very serious hardware. It is a start, and not a bad one at that,
Wi-fi doesn’t really have all that long of range. But I get your point. Once lower-power, meshing wi-fi is the norm, I think it will become somewhat of a standard.
Well… let’s say relatively long range. For consumer applications, it makes more sense to piggyback on existing WiFi infrastructure that have become ubiquitous, rather than building a parallel Zigbee network. Particularly, since the cost of embedding WiFi module into all kind of appliances is actually lower than Zigbee and it does not require any hubs to connect to the Internet. The only disadvantage of WiFi is relatively higher power requirements, although this is also being addressed.
TI CC3200 Low-power WiFi SoC can operate on 2 AA batteries for “up to a year” (depending on sleep/wake duty cycle of course): http://www.ti.com/tool/cc3200-launchxl
Due to sheer luck and finding a 1 level house needed due to some medical reasons, (short sale as well). I was able to find a 3000 sqft house here in VA. Very lucky and a great price (literally a fire sale). I have mounted my Router in the center in the attic. Gives me great reception all around the house. Same goes with the ST Hub.
I share a number of the concerns expressed in this thread but no need to repeat what I’ve written in my columns (such as http://rmf.vc/IEEENotScripted) nor memos from 20 years ago. At this point my goal is to understand what I can do with what is available from SmartThings – in a sense treating it like a Maker project rather than a consumer solution.
For those interested Echelon tried a similar approach 25 years ago and failed as a consumer product and shifted their business to embedded systems for commercial buildings.
@bobfrankston good write up in your column. I think 25 years ago that technology was just starting down this path. I know I was looking at solutions using 486sx based machines with A/D and D/A converters, and of course relays to make all the home automation work.
Obviously we have come a long way from those days, while the risks are still high for those entering the field, I think ST is a great example of what may be. There is still a lot of work that needs to be done here, and the interface is still a complex problem given the various styles and interfaces that they have to account for with their GUI. I believe the next big steps will finally take it out of the Maker category and move it into the Consumer level that they originally were wanting to accomplish
But so far we don’t even know what intelligence is! And compared to that dream, the today’s AI is just a giant toolbox for manipulating data. It’s still a human’s job to know how to use the toolbox and how to make sense of the data coming in and the data getting out. Yes, you can create a very sophisticated algorithm for solving a very specific problem way better than any human. Like flying an airplane, driving a car or trading stocks. BUT! 1) Some human has come up with that algorithm, not a machine. And when an algorithm runs across an unexpected set of inputs, it stumbles - very often quite spectacularly. 2) Anything less specific, especially when interacting with humans is involved, and we still have a looooong way to go. Probably decades before we come up with anything useful.