Singularity Alert: 3D Printing

[DontSayBanana]

Well, for starters, we need to talk about what we mean when we're talking about "great AI." Great graphics is pretty well understood to be improvements to resolution and scene lighting, but AI could mean just about anything from NPC pathing in a video game all the way to an evolving artificial personality.

I doubt a significant majority really wants a completely untethered artificial personality, since it's going to be capable of forming its own interpretations according to its own personally-formed opinions, and potentially countermanding instructions given to it.

A complete artificial personality would, by definition, be impossible to fully control.

[Admiral Yi]

The bottleneck here is computing power, not the state of the art. Gaming(and other) AI could probably be handled by a secondary GPU but developers aren't doing it because so few of their customers have a spare GPU attached to their computers. I imagine in the future high-end rigs will have a dedicated AI chip.

https://developer.nvidia.com/gpu-ai-path-finding

Really??  How does one go about writing an AI?  It's something i can't even fathom.  Surely you don't nest 3 trillion IF statements.

[Maximus]

The bottleneck here is computing power, not the state of the art. Gaming(and other) AI could probably be handled by a secondary GPU but developers aren't doing it because so few of their customers have a spare GPU attached to their computers. I imagine in the future high-end rigs will have a dedicated AI chip.

https://developer.nvidia.com/gpu-ai-path-finding

Really??  How does one go about writing an AI?  It's something i can't even fathom.  Surely you don't nest 3 trillion IF statements.

It's complex and depends very much on what you want to do(there's no universal AI, at least not yet), but basically you create a model of the system you want to work within, give all the costs and rewards numeric values, and then have a billion robotic economists figure out the cost/reward of every possible course of action and pick the best one.

That is the naive approach, there are optimizations to be made but that gets more complicated.

A good AI has to be able to choose the best model(or at least a "nearly best" one), update it as the situation changes, and choose optimal or nearly optimal courses of action, often within a fraction of a second.

[Iormlund]

As far as I know most game AI engines consist still of relatively simple expert systems (in essence the 3 trillion IF statements Yi metioned).

I'm not aware of things like neural networks or genetic algorithms having that much impact in the industry.

[Siege]

[Eddie Teach]

We're going to run into our biological limits sooner or later. It already takes people 15+ years of schooling to become productive members of society.

[Razgovory]

I guess the next logical step is for the graph to curve backwards as we invent time travel.

[Eddie Teach]

I don't think the size of the population is that great a proxy for the overall level of technology anyway.

[The Minsky Moment]

And industrial is more important in the grand scheme of things.

I disagree. Nothing changes the game in the life of a technology like when it goes from industrial to personal. Look at computers.

Every technology is different. 
With 3D printing there is a huge potential impact in terms of making highly customized manufacturing on a competitive cost basis with mass production, and having an enomous impact on supply chain management.
I don't think the consumer impact will be on that same level.   Yes it will be very convenient to order a product and have it "delivered" by an electronic instruction to one's personal 3D printer, but not a huge advantage over say having it fast printed in an Amazon warehouse and then delivered by speed drone.  It could have big impacts on education and obviously significant empowering effects on would be designers, inventors, etc.  But I wouldn't minimize the industrial impact in comparison.

[Siege]

The concept of ephemeralization:

"The ability of technological advancement to do more and more with less and less until eventually you can do everything with nothing."

(pioneered by Buckminster Fuller in his science-fiction novel 'Nine Chains to the Moon)

[Siege]

I'm sorry, this is too smart for me not to share it with you.

Quote from Curt Welch responding a singularity question:


I don't know what definition of "singularity" you are using, but it's clearly not the one I use.  To me, the singularity is the point in time where we first invent a machine that has all the basic physical and mental powers of a human, that we need for our economy.  That is, once we invent a machine, that can do all (or at least most) the jobs humans currently do in our economy.  What's happening in Mexico or in the third world countries has no bearing on when, and if, such technology is created so why you are talking about these other nations I have no idea.  It has nothing to do with the Singularity and more than what was happening in Africa on Dec 17, 1903 when powered controlled flight was first created by the Wright Brothers.

The singularity has been described in many different "magical" ways that are all just nonsense.  It's just a point in time where our machines we know how to built, become more intelligent than humans.  it will be no more magical than the point in time where the machines we have built, can run faster than human, or run faster than a horse, or fly higher than a bird.

Our machines passed human strength long ago with the steam age.  We all just accept that we are weak slow animals compared to the strength of a steam engine, or the speed of a jet aircraft.  Our machines passed human mental powers long ago as well.  They can compute and process information billions of times faster than a human can.  The google servers can read, and index web pages and images faster than any human can.  If we put every human on the planet to work trying to read and index all the web pages on the internet, they still wouldn't be able to do it as fast, or as well as the google machines can do it.  This fact that we are weak in strength, and weak in mental powers compared to our machines is not a problem.  We accept our limitations without a second thought.  We get the machines to do the things for us, that we can't do ourselves and we like having these machines to do this work for us.

There is one type of mental power we still have an edge on compared to the computers however. And that's our power for self directed learning.  We learn on our own, how to do new things, like build tools, and solve new types of problems. But that is the ONLY thing we still have over our machines.  Soon, our power to learn, which is the source of all our creativity that gives us the edge over our machines, will be lost as well.  I think this last piece of the puzzle will be cracked by 2020.

The "singularity" is the day we crack that last piece of the puzzle.  Life won't suddenly change anywhere because one guy, in some lab at google has cracked this last piece of the puzzle.  But it will open the door to what will become massive social change in the same way the invention of the airplane, or the telephone, opened the door to massive social change.

The technology is likely to be very expensive at first, so it may take a million dollars worth of computers and lots of electricity to duplicate the mental powers of even a single humans.  At that costs, most human jobs will still be safe.  But once the technology is invented, it will take off as fast as airplane technology exploded.

Once our machines can learn like humans, they will be better workers than we humans are at all major jobs.  Including the creation of art and all the creative jobs.  Even a computer with an average human IQ of 100 will surpass all other humans, because they won't need to sleep, or eat, or take vacations. They will work 24 hours a day, 365 days a year at whatever jobs we apply them to.  But not only do they have the edge in their work hours, they can share their learning with a simple download.  If we have a computer learning brain, and clone it into 10 robot bodies, and then each robot goes and learns something new for a week, at the end of the week, they can all share their new learned skills with each other, and end up with 10 weeks of learning, in that one week period.  By the ability to share learned knowledge with a simple download/merge, they will learn and advanced far faster than any 10 humans could.  These new machines will replace all human workers. They will take over all retail work.  They will take over the jobs of transportation.  They will do the manufacturing.  They will then move into the creative design and business management work. They will become the new entrepreneurs.

Most people don't grasp how close we are to replacing humans because none of our current machines ACTS like a human.  We seem to have very distinctive and different ways of acting, and doing things.  We are emotional, and oddly unpredictive in our behavior as well as creative and clever.  But all that is just the fallout of how self directed learning machines act.  And though we have examples of such behavior in the lab for simple domains, we don't have it yet for the complex high dimensional data enviornment of the real world yet.  But we are only one small algorithm away from having it, and those algorithms will be here real soon.  I'm one of the many people working on creating them and I have a good sense of how close we are.  It's a puzzle I've spent 30 years working on, and I've see how much progress has happened in the past 30 years because I've been closely following it.

Once we have these algorithms working, our robots will for the first time, start to act like real living animals, and humans.  And once people see the robots actually acting, and learning, like an animal, they will finally grasp just how truly close we are to duplicating human skills in our machines.  Most people need to see these things before they can understand them, and as such, until we have the thing for them to look at, they will keep suggesting "we aren't even close" when in fact we are one step away.

The same was true for the airplane.  A large number of people assumed heavier than air controlled flight was impossible.  They said it was impossible even after the Wright Brothers had made their flight.  They said the Wright Brothers were obviously lying.  But once they saw a plane flying in circles with their own eyes, they understood.  A large percentage of the population just works that way.  They need to see things before they can understand them, and as such, they have no real ability to predict the future.  But they don't have any problem understanding it AFTER THE FACT.  These human-like robots are going to take many people by surprise, even though we have all told you they were coming.

[The Minsky Moment]

I don't know what definition of "singularity" you are using, but it's clearly not the one I use.  To me, the singularity is the point in time where we first invent a machine that has all the basic physical and mental powers of a human, that we need for our economy. 

Once that happens, whatever constitutes the economy changes.  It's a dynamic process.
Think about a pre-modern or even early industrial economy: the basic production tasks are planting, reaping, sowing, transporting, and farbicating basic goods like clothing, furniture, horseshoes, etc.   Things that machines can all do now and have been able to do for decades now.

Once you begin to deploy machines to do such tasks "our economy" enlarges to include ever more varied and complex tasks and concepts and so the bar moves out further.

[Siege]

I don't know what definition of "singularity" you are using, but it's clearly not the one I use.  To me, the singularity is the point in time where we first invent a machine that has all the basic physical and mental powers of a human, that we need for our economy. 

Once that happens, whatever constitutes the economy changes.  It's a dynamic process.
Think about a pre-modern or even early industrial economy: the basic production tasks are planting, reaping, sowing, transporting, and farbicating basic goods like clothing, furniture, horseshoes, etc.   Things that machines can all do now and have been able to do for decades now.

Once you begin to deploy machines to do such tasks "our economy" enlarges to include ever more varied and complex tasks and concepts and so the bar moves out further.

So, are you saying the impact in the economy cannot be predicted because the bar is going to move farther out?

I myself have many doubts about the singularity as proposed by Kurzweil, mainly the part about strong AIs and sentient machines.
However I think he got right the part about exponential technological grow, which means our economy will be impacted by an increase in the use of robotics with a large potential for massive unemployment. If this is how it goes down, then the argument can be made that the free-market forces to promote technological development will end up forcing a welfare state as the only viable solution to massive unemployment produced by technological developments.

I don't know. I think only people with technological degrees will benefit from all this.
And the capitalists pigs, of course, who will own the new push-button factories.

[The Minsky Moment]

So, are you saying the impact in the economy cannot be predicted because the bar is going to move farther out?

An economy is just the sum of activities that human beings engage in that are deemed by them to be productive (in a market economy because they implicate market transactions).  So it is inherently dynamic and to a certain extent socially determined.

Imagine going back to the 1810s or 1820s in America or Western Europe at a time where over 90% of people were engaged in farming.  Now imagine telling these people that in a 100 years that proportion would in some places dip below 3%.  I.e. pretty much the only thing that people thought of as productive work was going to virtually disappear as a form of employment.  The reaction would be dismay and confusion, because it would be impossible to imagine what other forms of work could possibly arise to replace virtually every job in existence in such a short (historically speaking) period of time.

Another little thought experiment back in 1810.  Imagine the village of farmers you are visiting is having a festival day; in a field some young men are playing a traditional ball-game, quite literally as a past-time.  Now imagine explaining how in the future, not only will one will be able have a "job" consisting of playing this game - but that some of those holding such jobs will be among the wealthiest and best paid people in the country, and that additional jobs will exist in the tens thousands of people to coach and manage, provide instruction, operate and capitalize joint-stock companies that employ such people, convey these players around the country for their exhibitions of skill, write about and talk about their games in newspapers and other forms of communication they would find unimaginable, and even give advice to people on how to construct imaginary groups of such players.  They wouldn't just think you were insane.  They would simply be unable to understand what you were talking about - there would be no frame of reference in their minds to understand it.

That is position we are in now thinking about the jobs a hundred or so years from now.

[Siege]

That is position we are in now thinking about the jobs a hundred or so years from now.

Good point. However the difference is that we, as 21st century people, at least have the perception that technological developments can change our economical landscape beyond what we can imagine. Or do we? I don't know. Kurzweil says the we, as humans, think of technological development in lineal terms. We think technology is gonna keep improving, but in general terms is gonna be at a predictable development level. I.e. cars are going to be better, more efficient, more automations, but they are going to still be cars; TV sets are going to be better, super HD, 3D, but still TV; and computers are going to be faster, and smaller, but still gonna be recognizable as a computer.

Reality is, if the guys experimenting with wearable computers and visual interfaces succeed, and nanotech allows to build nanochips, everything is going to be different. Cars are going to be self-drivable, and by their 3rd Gen they will probably not even have steering wheels, with their interior radically redesigned. TVs are gonna go to the extremes, with the sincret-walls having an entire side of room as a wall-sized TV, and smart glasses/contact lenses having a simulated HD screen in front of your eyes. Two of the currently developing technologies that I think will really revolutionize our world are the IoT (internet of things) and enhanced reality. The IoT will put every single piece of equipment online, being able to relate information to us through smart phones/glasses. Enhance reality will give us stats and information on everything we look at. Still, I think long term the core technologies that will be more life changing are the so-called GNR revolution, Genetics, Nanotechnology, and Robotics.

Anyway, I digress. I see your point, and I agree eventually our economical and social system will adapt to whatever changes technology brings in the future. Nevertheless, I fear mass unemployment could be in the cards, and people might have to develop new skills to survive. Perhaps we are all going to end up working 20 hours a week, online from home, in some academic field unrelated to actual physical production of material goods. Or, maybe 3D printing will end with the concept of the factory. At least for consumer goods. Regardless, the future will be amazing beyond our dreams. If we don't nuke ourselves back to the stone age.