a thoughtful web.
Good ideas and conversation. No ads, no tracking.   Login or Take a Tour!
comment by theadvancedapes
theadvancedapes  ·  3700 days ago  ·  link  ·    ·  parent  ·  post: The Road to Superintelligence

    Is it really omniscience if I have a personal assistant that provides information for me? Technically, Siri and Wikipedia does just that

Yes, but Siri is useless, and we don't have a knowledge base on the Internet to make sense of everything on Wikipedia for us. Siri and Wikipedia are primitive precursors of what's to come. And this isn't about us being omniscient. This is about us interacting with an omniscient knowledge base.

    I'm not convinced we'll soon be creating everything we want from our homes.

You won't necessarily be creating everything you want, but you will be able to order anything you want from the Internet and find it printed out for you at home. All for next to nothing. There is no reason to suspect that you won't have torrents for things like violins, bikes, or even cars.

    The description could just as well be about the internet itself, with its improved connectivity, abundance and education potential.

Internet is the medium that we are building the global brain with. But the Internet is not yet fully mature, as it is not yet a single information processing system for the entirety of human interaction, as well as human transportation and energy use. But that is where we are heading. The Internet by itself cannot provide material abundance, we need to build ICT that will allow us to integrate technologies to get abundance for all. Same goes for connectivity and education. Our Internet isn't equipped with advanced AI software helpers or semantic knowledges bases, etc.





veen  ·  3700 days ago  ·  link  ·  

    And this isn't about us being omniscient. This is about us interacting with an omniscient knowledge base.

How would you bridge the gap between a knowledge system that holds information and a knowledge system that can determine meaning out of the information? There's Google trying to create something similar to an omniscient base with their Knowledge Graph and Google Now technology, giving me information before I know I want it. But it is merely based on extrapolation and assumptions. IBM's Watson is the same thing. It can already do calculations faster than humans, but it's still programmed to do so. The technology doesn't know why it does what it does, it can't give meaning, and in my mind an omniscient knowledge base worthy of the term singularity should be able to do so. That's a really big step AI needs to take that I don't think it will.

    All for next to nothing.

Maybe for plastic, but if we have a completely subdivided manufacturing economy, the only thing worth making money off is the materials. Higher demands create higher prices, and the transportation costs will rise dramatically. Materials becoming dirt cheap seems like an assumption too easily made.

    as it is not yet a single information processing system for the entirety of human interaction, as well as human transportation and energy use. But that is where we are heading.

I feel like I'm missing some vital information as I don't really get what you're trying to say here. Do you want every interaction to be connected to this network, the global brain? That the IT network will level the playing field, connecting everything and everyone and thus create abundance for all? Reminds me of Thomas Friedman's claim that the world is becoming flat (meaning no transportation costs, making place irrelevant in an infinitely connected world). This article by Richard Florida is a great response to why that line of thought is wrong.

theadvancedapes  ·  3699 days ago  ·  link  ·  

    How would you bridge the gap between a knowledge system that holds information and a knowledge system that can determine meaning out of the information?

We will undoubtedly get much smarter. I didn't go into it in any detail, but you'll see one of the mechanisms for omniscience is internal computing. That would mean that we would essentially upgrade our neocortex. I feel very strongly that we will massively upgrade our neocortex by connecting it to the cloud, there are a lot of evolutionary trends in our history that suggest major transitions (such as a global brain) would also result in a extension of our neocortex. This would obviously be a radically leap though. Creates a new world. But global brain will always be more powerful and knowledgeable than any one of us.

    IBM's Watson is the same thing. It can already do calculations faster than humans, but it's still programmed to do so. The technology doesn't know why it does what it does, it can't give meaning, and in my mind an omniscient knowledge base worthy of the term singularity should be able to do so. That's a really big step AI needs to take that I don't think it will.

Yes, I agree with all of that, but I believe we will figure out how to make sense of all human knowledge. It will just take a massive leap in our software capabilities. Being incredulous isn't evidence that we won't do that. I suspect that the biggest improvements in software will come when we completely understand the human brain. That would make sense, since it is human thinking we are trying to duplicate.

    Maybe for plastic

When you manufacture on the atomic scale it's all the same.

    the transportation costs will rise dramatically

No. Transportation is going to be free and infinite. We also won't be driving any vehicles, they will be driven by AI.

    I feel like I'm missing some vital information as I don't really get what you're trying to say here.

Industrial Internet. Here's more.

    Do you want every interaction to be connected to this network, the global brain?

It is not about want. It's about the system-level trends. Everyone will be using the Internet for every experience in their life by 2040 (with the aid of full-immersion virtual reality, wearable and than internal computing and related technologies). Also the whole environment will be online. This stuff will happen.

    That the IT network will level the playing field, connecting everything and everyone and thus create abundance for all?

Pretty much.

    This article by Richard Florida is a great response to why that line of thought is wrong

I strongly disagree.

kleinbl00  ·  3699 days ago  ·  link  ·  

    There is no reason to suspect that you won't have torrents for things like violins, bikes, or even cars.

This is a common fallacy amongst fans of 3d Printing. I'm a gonna go ahead and call it what it is - Cheez Whiz Engineering.

You list three things. Let's talk about what goes into them.

1) A violin. It's a resonating box with resonators on top. Resonation is a function of density and tensile strength, those densities and tensile strengths being 3-cimensional matrices in tension, 3-dimensional matrices in compression. In order to "make" a violin I need one non-uniform, non-linear material graph for the body. I need another non-uniform, non-linear material graph for the neck. I need another non-linear, non-uniform material graph for the bridge and yet another for the strings. I can build the tuning pegs out of Cheez Whiz, but the sound of a violin is entirely dependent on the resonant properties of those four deep non-linearities.

2) A bicycle. Let's just go to the frame. Suppose I make it out of chrome-moly. This is an anoxic carbon steel blend forged in one atmosphere, tempered in another, joined in another. My fabrication processes are well-understood but they do not lend themselves to thermoplastics. There are parts of a bicycle that lend themselves to Cheez Whiz, but they're the internal guts of the shifters and the Cheez Whiz they lend themselves to is sintered metal powder. Not exactly something user-manipulable on 120W.

3) A car. Let's simplify and go with something stupid simple like "a brake disc." Not the fluid, not the pads, not the bearings, not the studs; those are all deep chemistry and metallurgy. A brake disc is forged in the fires of Hell to do one thing and do it well. It's a multi-tempered multi-pass multi-machined component reliant on many different source materials combined in concert in extreme processing in order to produce one dumb, homogenous part that could be drawn as an STL file by any mook. Make it out of Cheez Whiz, though, and…

Here's the problem with 3d printing. It presumes that all things are made of thermoplastic. Granted - there are things that can be made from thermoplastic. Dice. Watchbands. iPod docks (minus the circuitry). Flowerpots. Amusing 3d Solids. They're lowest-common-denominator strength, though, and usually highest-common-denominator cost. There are very 3d printed things in the world that can't be made radically cheaper via lost wax or injection molding.

So if there's nothing in your life that needs any material strength, you can totally 3d print your entire life. But if you actually need functional goods, you shall remain dependent on the manufacturing apparatus arrayed around you. Yeah, you can 3d-print a bottle opener. it'll cost you 3-4$ in materials. Or, you can go to the store and buy a better bottle opener for 79 cents. They're much easier to stamp out of steel, and they work way better.

That's a bottle opener. You don't want to imagine a Cheez Whiz bicycle.

Yeah, you can print a violin. and if you like listening to Cheez Whiz, that's a bright, bright future.

theadvancedapes  ·  3699 days ago  ·  link  ·  

I think this is perhaps your crucial point where we can find agreement:

    But if you actually need functional goods, you shall remain dependent on the manufacturing apparatus arrayed around you.

I agree that for many things you will not necessarily be able to print them out directly from your home. But the manufacturing apparatus around you will be able to. Hospitals will have sophisticated 3D printers designed specifically for printing out organs. Automobile manufacturers will have essentially replaced their assembly lines with 3D printers, etc. All of this will dramatically reduce the cost of material goods.

kleinbl00  ·  3699 days ago  ·  link  ·  

Not really, no. Your counter-argument is "contractor-grade Cheez Whiz."

3d printing is integrated, incremental materials deposition. That's the definition. Your "cell" is your printing resolution. Which works fine for amorphous materials, but runs into real problems as soon as you're dealing with anything beholden to materials science.

Take a fishing rod. We'll skip the rings and cork and shit, we'll just talk about a long piece of fiberglass. It gets its strength from a bundle of glass fibers that run longitudinally and a matrix of polymer that binds them together. It's a "wood" analog - the fiberglass is a crude imitation of bamboo, essentially.

The glass fibers are microscopic in one dimension and macroscopic in another, running to "microns" and "multiple feet" respectively. The effective manufacturing methods for fishing rods involve aligning bundles of glass fiber on a jig, depositing matrix (spray or injection, not sure), using vacuum to remove air pockets, and then allowing the matrix to cure before polishing the assembly. You now have a stick of plastic and glass that performs better than a piece of bamboo, but not quite as well as hand-formed cane.

Let's take this process and turn it into 3D printing. Now our glass particles are microscopic in all dimensions. We have no stress matrix. Effectively, we've got a pile of sandpaper without the paper - glass particles in glue. We can heat-cure the assembly… but now instead of having a sturdy and resilient pole, you've got a glass rod. Or, more specifically, a glass-and-glue rod.

That's 3D printing - integrate down to zero, find the ideal particle for your nozzle size, and repeat it volumetrically. It's like making an English longbow by taking the yew tree, feeding it to the wood chipper, rendering it to sawdust and then forming the paste into an arc. You can't make a decent bow out of MDF, let alone plywood. Yeah, you can add all sorts of mathematical filigree but in the end, your maximum imodulus is your particle size. If all you've got are 1x4 legos, you're going to have a rough time building a broomstick.

A "How it's Made" marathon might do you some good - we have hundreds of different manufacturing processes because we're efficient, not the other way 'round.

As to the 3D printed organs:

The technology they're leveraging is based upon a handful of patents developed by the University of Washington in the mid '90s. I did a lot of post-grad work with them. Basically, Mat Sci figured out a way to create fibers so small that the body didn't reject them but so inert that they didn't cause mesothelioma like asbestos. They kind of work as a matrix for stem cell growth - we'd spray 'em down electrophoretically and then put some cultured tissue on 'em and they sort of grow in the shape of your matrix. At the time we were looking at wound care because skin tissue was more easily cultured and easier to acquire than, say, lung tissue but the principle is the same - make a "something" and coax cells into growing on it.

Yeah, you can "3D print" that matrix, after a fashion. We called it "spraying" back in '99. Of course, that's back when our 3D printer was called the STL lab and nobody cared… nowadays, you say you're "3D printing" and suddenly The Economist wants to come shoot videos.

Make no mistake - the tech hasn't changed. It was a questionable tech spinoff 15 years ago. The Vacanti Mouse was 17 years ago… how much has it changed the world?