Insights about the Semantic Web
Some insights about the Semantic Web:
As I understand it, the Semantic web is basically an internet where for every page on the web, there's also a crapload of additional data layered on top of it that provides context and relevance. In fact, there could be waaaay more context and relevance data in a webpage than actual data on the page.
This is a lot like memories in your head: Every memory, no matter how insignificant, is connected in some way with a family of concepts, which is what you use to find the memory when you want to use it. Say you're thinking about summer days, and contained in summer days are a bunch of memories of actual summer days, as well as stuff like "hot" and "dry" and "muggy" and "nice smells" and whatever else you associate with your experience of "summer days". Each of these concepts leads to their OWN raft of memories and after a few seconds of thinking about "Summer days" you might find yourself drifting off to memories of apple orchards or water slides or whatever else you're interested about in connection with the original memory. Building the web like this would make it a lot easier to find stuff you're looking for, because the concepts would chain together in logical structures according to relevance.
The thing is, this contextual info can actually be way more information than the actual objects it carries, so efficiency really matters. Consolidating memories happens all the time, and the context web is super-efficient. But basically your brain has to be more than twice as active than it would normally take to process the actual data held in it. It has a huge informational overhead for every actual memory you've got stuck inside your cranium. This is probably why attempts at building artificial intelligence have failed, because the context web has been too thin for each one. A computer contemplates a block, but it doesn't get the flood of contextual info that would inform a baby contemplating the same block. Hell, it even has trouble identifying the block, so we know we've got a ways to go here.
If it DID get a context flood, you would see the computer playing with the block in the same way a baby might, because the computer would have so much contextual info associated with the block that it wouldn't know what to do with it all, so it might just fiddle around with it until something interesting happened. This is why babies play with toys: they are looking for interesting permutations of their contextual web by putting the object into new relationships. If a computer started spontaneously playing with the toy in the same way, you'd know that you were close to something that could think, if you could only scale that introductory context web up until the computer had advanced context webs for extreme concepts like words and stock options.
Artificial intelligence fails because the context web density and efficiency for any system that's ever been made has fallen far short of true intelligence. The context web needs to be geometically more dense than the concepts embedded in it, and computers just aren't up to the task yet. Your brain makes it look easy, but it has all sorts of cool tricks that we know nothing about. But I digress.
The semantic web is an attempt to bring the same kind of thing to the internet, so that companies like Google will eventually be obsolete, or mostly obsolete. An internet where you put in a concept like "rubber" and you get back a torrent of other concepts from tires to latex fetishism, then you pick the one you want to follow and off you go until you find what you're looking for.
Google tries to do this, but if fails miserably because it tries to shortcut the task to one inference and an instant result. And, frighteningly, the web is just too damn big. Google could choke itself trying to provide contextual info on every page inside it. The semantic web is an attempt to offload the whole task onto the web itself so that each page knows what it's about and what it relates to and can provide that info without Google trying to do everything in one location.
So basically we're trying to turn the internet into a big brain so that we can find stuff easier. When it gets done, creativity and productivity will explode because you'll be able to put in something like "asian fighting kite designs" and miraculously discover that there's a little old lady six blocks over from your house who is ALSO interested in asian fighting kite designs and would be happy to show you her collection and talk about it with you. So everybody becomes more productive at everything.
I think it's a pretty cool vision.
As I understand it, the Semantic web is basically an internet where for every page on the web, there's also a crapload of additional data layered on top of it that provides context and relevance. In fact, there could be waaaay more context and relevance data in a webpage than actual data on the page.
This is a lot like memories in your head: Every memory, no matter how insignificant, is connected in some way with a family of concepts, which is what you use to find the memory when you want to use it. Say you're thinking about summer days, and contained in summer days are a bunch of memories of actual summer days, as well as stuff like "hot" and "dry" and "muggy" and "nice smells" and whatever else you associate with your experience of "summer days". Each of these concepts leads to their OWN raft of memories and after a few seconds of thinking about "Summer days" you might find yourself drifting off to memories of apple orchards or water slides or whatever else you're interested about in connection with the original memory. Building the web like this would make it a lot easier to find stuff you're looking for, because the concepts would chain together in logical structures according to relevance.
The thing is, this contextual info can actually be way more information than the actual objects it carries, so efficiency really matters. Consolidating memories happens all the time, and the context web is super-efficient. But basically your brain has to be more than twice as active than it would normally take to process the actual data held in it. It has a huge informational overhead for every actual memory you've got stuck inside your cranium. This is probably why attempts at building artificial intelligence have failed, because the context web has been too thin for each one. A computer contemplates a block, but it doesn't get the flood of contextual info that would inform a baby contemplating the same block. Hell, it even has trouble identifying the block, so we know we've got a ways to go here.
If it DID get a context flood, you would see the computer playing with the block in the same way a baby might, because the computer would have so much contextual info associated with the block that it wouldn't know what to do with it all, so it might just fiddle around with it until something interesting happened. This is why babies play with toys: they are looking for interesting permutations of their contextual web by putting the object into new relationships. If a computer started spontaneously playing with the toy in the same way, you'd know that you were close to something that could think, if you could only scale that introductory context web up until the computer had advanced context webs for extreme concepts like words and stock options.
Artificial intelligence fails because the context web density and efficiency for any system that's ever been made has fallen far short of true intelligence. The context web needs to be geometically more dense than the concepts embedded in it, and computers just aren't up to the task yet. Your brain makes it look easy, but it has all sorts of cool tricks that we know nothing about. But I digress.
The semantic web is an attempt to bring the same kind of thing to the internet, so that companies like Google will eventually be obsolete, or mostly obsolete. An internet where you put in a concept like "rubber" and you get back a torrent of other concepts from tires to latex fetishism, then you pick the one you want to follow and off you go until you find what you're looking for.
Google tries to do this, but if fails miserably because it tries to shortcut the task to one inference and an instant result. And, frighteningly, the web is just too damn big. Google could choke itself trying to provide contextual info on every page inside it. The semantic web is an attempt to offload the whole task onto the web itself so that each page knows what it's about and what it relates to and can provide that info without Google trying to do everything in one location.
So basically we're trying to turn the internet into a big brain so that we can find stuff easier. When it gets done, creativity and productivity will explode because you'll be able to put in something like "asian fighting kite designs" and miraculously discover that there's a little old lady six blocks over from your house who is ALSO interested in asian fighting kite designs and would be happy to show you her collection and talk about it with you. So everybody becomes more productive at everything.
I think it's a pretty cool vision.