I had a discussion with a friend recently about ways to fix the economy, and while I had a few ideas I didn't have a solution for unemployment. Now I've thought it over a bit and I wonder if software can't help out.
Here's the main problem: to increase the economic output of the world economy, high-tech manufacturing is a must. This includes local fabrication shops, black factories (no lights inside, because its all autonomous), nanotech facilities, what have you. All of this has the large problem of requiring a great deal of education to run, and the preponderance of the unemployed won't have the educational background for any of these jobs. I'm thinking that appropriate software could bridge the gag. Rather than trying to develop software to make systems entirely autonomous (a very difficult thing to do), it may be easier to use the software to simplify the job down to something that can be managed with a minimal education.
This calls for a tall software stack between the user and the problem set, and any complex system like that will come with its own repair and upkeep problems. Security becomes a giant problem, and there is no way to fully "wrap" the problem with automated systems: you will still need plenty of people who know how to fix and maintain the software that's already in place. But still, I think you would end up with a vastly expanded industrial base if you can optimize computer/industrial systems for ease of learning and use.
Welcome! These forums will be deactivated by the end of this year. The conversation continues in a new morph over on Discord! Please join us there for a more active conversation and the occasional opportunity to ask developers questions directly! Go to the PS+ Discord Server.
Solving employment problems with software
Tue, 2011-08-30 14:47
#1
Covariant
Solving employment problems with software
Wed, 2011-08-31 05:11
#2
Arenamontanus
Re: Solving employment problems with software
I have heard that this is what Walmart and some other retail stores use for their teller systems: by making a simple iconographic interface they can hire people without expensive reading/language skills.
In one of my ongoing EP campaigns this approach is used on a giant scale: http://eclipsephase.com/build-nation - Tanzania and a microcorp are trying to do massive nanoconstruction using millions of infugee citizens playing something like Farmville. If it works, then the microcorp has proven their software stack.
In reality, this kind of software stacking only works as long as AI is more expensive than the human plus software. Once pattern recognition and language understanding become good enough there is likely little need for most humans in such systems. Manufacturing jobs are not the future for humans.
http://www.scdigest.com/assets/newsViews/08-06-12-2.php
—
Wed, 2011-08-31 05:45
#3
Extrasolar Angel
Re: Solving employment problems with software
The problem is that you don't need to employ people once such software is introduced.
Unemployment today isn't connected to lack of education. Unemployed of today (including me for over 3 years) are not underqualified but overqualified and overeducated.
We have passed the threshold of "peak employment" I am afraid, and newer technological solutions simply reduce the need for people to be employed, rather than providing new jobs.
—
[I]Raise your hands to the sky and break the chains. With transhumanism we can smash the matriarchy together.[/i]
Wed, 2011-08-31 06:52
#4
Xagroth
Re: Solving employment problems with software
Add to that the simple yet usually ignored fact that unemployed people won't spend money, which leads to a diminishment on sales, and you have the why of economic collapse under an economic system that requires constant buy from the people to keep growing.
In short: if you make all bussiness to require no people so you don't need to pay them... then in a short time only those bussiness will have money, and everybody else would have turned to a trade system out of sheer necessity.
Wed, 2011-08-31 09:36
#5
Extrasolar Angel
Re: Solving employment problems with software
Indeed the current situation seems hopeless. Unless massive changes to regulations are made forcing essentially to stop eradication of jobs(including by technology), the only hope will be massive collapse or social upheaval.
—
[I]Raise your hands to the sky and break the chains. With transhumanism we can smash the matriarchy together.[/i]
Wed, 2011-08-31 12:03
#6
Arenamontanus
Re: Solving employment problems with software
You are forgetting that most jobs are in the service economy, where automation does not bite that much yet. Unemployment is a rather complicated beast.
Also, there might be plenty of profit in a business-to-business economy. Remember that companies are consumers too.
In the very long run when AI can do anything a normal or even skilled person can do, it looks like the only way to get paid is simply to own some stock in the AI-run corps and live of the dividends. The growth of an AI economy appears to be pretty explosive, so those dividends might actually be pretty high. So if rising AI skill looks plausible it would be rational for governments to turn their investments into welfare systems into stocks, and then either distribute them or the resulting dividends to people in the system. Of course, some governments might be more interested in getting rid of their citizens...
—
Wed, 2011-08-31 12:53
#7
Arenamontanus
Re: Solving employment problems with software
This hilarious little clip seems relevant:
http://boingboing.net/2011/08/31/with-computers-doing-the-thinking-the-e...
—
Wed, 2011-08-31 15:09
#8
750
Re: Solving employment problems with software
It depends on how one define service i guess.
Hell, automation have even reached lawyers. Now programs can sort thru large piles of documents in search of various topics (and not just direct words or synonyms, but context even).
And there are attempts at automating burger joints or allowing customers buying clothes online to actually see how well a item will fit thanks to adaptive manikins and AR.
Thu, 2011-09-01 00:55
#9
Covariant
Re: Solving employment problems with software
Automation has indeed reached the legal system. It's also taking chunks out of the financial industry and the medical industry. But then, software has taken a huge chunk out of its own industry: the hallowed automated code robot of yesterday's science fiction is alive and well today in the form of advanced software compilers.
Software takes chunks out of every industry, and will continue to do so until that mythical day that we create strong ai. Software development itself is protected from the onrushing tide of progress because it is inherently a complex process. It is at best an almost decomposable problem, and is ai-complete. So I ask myself "how do I frame employment industries to give them the same attribute?" More simply put: "How the do I keep the robot from taking my damn job?"
I have a number of ideas for that, but I'm also sure that this is a problem that managerial scientists, economists, and ever blooming hypercorp has come up with an answer for. #CharlesStross may have had the time to check on all the newest and deepest thinking on the subject, but I haven't. I'm just trying to come up with iP(o,a)d apps that allow people to do the job of a small army of worker drones for whatever they work on.
Speaking of which, apparently those swarm combat drones we've had flying around poor parts of the world blowing shit up have inspired a company in Iowa to use the same tech for farm work.
Thu, 2011-09-01 06:13
#10
Arenamontanus
Re: Solving employment problems with software
Automation can be complement or substitute to human labour: we don't have fewer lawyers due to software, just lawyers who can more easily bring up information and do more work per person - the software complements the lawyers. Other jobs get substituted away, like car assembly line workers. Substitution does not automatically make people unemployed, but typically means there are more work at a higher abstraction level - which is bad news for unskilled people (for example, in a mine I recently visited people sat in front of computers overseeing drilling robots rather than drilling themselves).
Weak AI only substitutes a few kinds of simple jobs, while strong AI might in principle substitute nearly any job. There are a few jobs that might not be substitutable at all, because they have symbolic or reputation aspects: think of the Pope. Or a real human (and hence expensive) butler. It is not how they do their jobs that really matters, it is who they are.
If you make employment industries as hard to change as software, then I bet some other industries that are easier to upgrade will flank them. Case in point: the music industry had developed an elaborate structure that was very hard to change and guaranteed everybody in it employment. Then the Internet and Apple came along, and now it is increasingly at the mercy of software companies. On the other hand, if an industry becomes very agile it will tend to subsume other industries, like software or gas stations.
The hardness of software intrigues me. Is it just because we humans are bad at writing it, or is there a real complexity based difficulty? If it is just human thinking, then we should expect AI to become rapidly better at making it in the future. But if it is actually hard for any intelligence to make software, then it makes rapid singularities much harder and slower. And the exact ways in which it is hard has big effects on how the singularity plays out - would emerging superintelligences need a lot of computing power, would they need a diversity of approaches, is it more a matter of gathering lots of data?
For competitiveness, I hope they can run them using fewer people. Military drones have 150 persons in the total team running each drone.
—
Thu, 2011-09-01 09:49
#11
Xagroth
Re: Solving employment problems with software
It's not the same thing: in a farm you risk less running into unforseen problems, and the drones are not expected to suffer as much wear and tear as combat drones.
Anyway, my point was that the "traditional" economies presented in Sunward are like terminal patients on life support... unless they remove the other options, of course, in which case they become the only option.
Thu, 2011-09-01 11:26
#12
Covariant
Re: Solving employment problems with software
This is what I am looking for, actually. Good software allows a single person to do more work than they could before and completely removes certain types of simple jobs.
Agreed, software brings jobs to a higher level of abstraction, and the people who get paid the most these days are those who can handle ever-higher levels of abstraction. The true challenge is designing software that allows unskilled people to do jobs they otherwise couldn't. For instance pilots (very skilled people) are utterly reliant on a system of checklists when things go wrong in flight. Airplanes are so complicated that no person will ever understand the whole thing, so there are mountains of what-if books kept on electronic file for the pilots to whip out in times of emergency. Now the pilots can take on the job not only of flying the plane but temporarily taking on the job of a thousand other specialists by following what is essentially a highly sophisticated help file.
I'm not looking for a protectionist answer to employment industries, I'm looking for sets of related tasks that aren't reducible to software. Sales, for instance, is a nebulously defined industry that is ai-complete; there are just too many fluid aspects to it that rely on being human.
It is inherently hard. Humans suffer from limits on how much complexity we can deal with, the old short-term memory being 7±2
chunks of information problem, but a more intelligent system will still have some finite limit on chunks it can handle while the possible complexity of software systems is infinite. There are also bottlenecks and black swan effects. Charles Babbage invented the first computer and Ada Lovelace invented the first program, but they were choked by the bottleneck of size and power in their mechanical computing machines. Things slowly got better with vacuum tubes and the like until the black swan invention of the transistor allowed us to massively increase the amount of computing material we could cram into a given area.
However, this also means that there isn't any good way to predict the path of a singularity. Maybe we will discover some quantum optical effect that allows neural architecture to run more efficiently by an order of magnitude and it self-organizes in ways that we haven't discovered algorithms for yet. Maybe we will discover that our networking algorithms for handling mobile device meshes has emergent properties that behave intelligently but at a level of abstraction that we don't understand until it's too late. Maybe we will have to put exponentially more effort and energy into every addition to intelligence and the singularity is choked by lack of access to cheap energy. It'll be interesting to watch any which way it goes.
Thu, 2011-09-08 19:18
#13
Covariant
Re: Solving employment problems with software
http://ieet.org/index.php/IEET/more/tarifi20110908
An interesting model of economics to get around the dramatic increase in available labor caused by automation.
Fri, 2011-09-09 05:19
#14
Arenamontanus
Re: Solving employment problems with software
Hmm, besides the big problem of how to implement something like that if you don't start from scratch, it seems to get in trouble when you have population inflation. If I fork, each of my forks now have a right to the basic income guarantee.
The government must also track the actual resource state accurately, but it is not clear what incentives there are for it to do so other than general niceness. Especially since economic states can compete with each other for the best people and projects by making different adjustments. As one of the commenters noted, this system probably gets into trouble with rational ignorance among voters.
—
Fri, 2011-09-09 05:39
#15
750
Re: Solving employment problems with software
The more i learn about our inherent mental biases, the more i worry how far up shit creek we really are as a planet.
Fri, 2011-09-09 07:09
#16
Arenamontanus
Re: Solving employment problems with software
The biases usually make things worse (but not always, we have altruistic biases too). But even very rational beings would have trouble with these issues. If we all want what is good for us but need to coordinate to get it, yet the coordination is too hard to do or game-theoretically unstable, then we will not be very good at achieving it even if we are rational about it.
Economics is about how to allocate scarce things. If mental power is no longer scarce, then the basis for the economy (and the struggles) will shift towards whatever it is that is scarce. It might be something as intangible as social status, which has the anti-inflationary effect that it is purely positional but also has a very nasty zero-sum game aspect.
—
Fri, 2011-09-09 08:33
#17
750
Re: Solving employment problems with software
Heh, i have a separate issue with game theory. This because the guy that did the more elaborate work on the concept may have introduced paranoia into the theories that are not there in real life.
Too me, game theory is second only to neo-classical economics in clinging to axioms that are either questionable or outright wrong.
Fri, 2011-09-09 09:03
#18
Arenamontanus
Re: Solving employment problems with software
You mean the von Neumann–Morgenstern axioms? People who actually work in the field recognize them for what they are, and commonly explore games and agents that do not satisfy them. Sadly, that is research few people outside the field hear about. A bit like how behavioural economics did away with Homo economicus years ago and physicists do not always assume everything is frictionless.
(Coincidentally, some of us were dissing the axioms over dinner yesterday. Humans clearly do not have transitive preferences. But even for agents that obey the axioms friendly superintelligence seems very hard to achieve...)
—
Fri, 2011-09-09 09:27
#19
750
Re: Solving employment problems with software
Sorry, nothing as detailed. Only know that when either the prisoners dilemma, or something similar, was tested on the RAND secretaries back in the day they behaved quite differently from what the game suggested they should do. And yet it continued to be used rather then set aside as a similar theory in physics would have ended up.
Fri, 2011-09-09 10:25
#20
Arenamontanus
Re: Solving employment problems with software
Hmm, if you demanded that strictness you would not be able to claim humans have languages, emotions or societies. And you should have rejected the claims that stars were distant suns rather than lights on a sphere, since you could not see them move relative to each other over a year. There are some rather fundamental differences between studying systems composed of simple (atoms) and complex (humans) entities.
The prisoners dilemma and how people behave in it has turned out to be a very rich field. The fact that people do not behave as perfect rational agents is a very interesting fact on its own. And the fact that nuclear superpowers composed of these agents do behave much more like it is another even more surprising fact. And the rich dynamics of the iterated evolutionary prisoners dilemma seems to be telling us quite a lot about societies. It is just that first year economics rarely has the time to get into the real stuff.
—
Fri, 2011-09-09 11:01
#21
Covariant
Re: Solving employment problems with software
Yea, I'm gonna go ahead and say that game theory is one of the best new fields of mathematics. Games like the Prisoner's Dilemma may not model the full complexity of human interaction since it doesn't take into account the relationships between the prisoners, but that's not what it's for. It has amazing uses for artificial agents, and wonderfully models the behavior of agents when there is no communication path between them other than their actions. For human psychology, the way that we fail to perform as expected by game theory isn't a weakness in the theory, it's an interesting indication of how we diverge from purely rational beings.
Mon, 2011-09-12 04:47
#22
Arenamontanus
Re: Solving employment problems with software
Or even that the economical model of rationality is flawed and there is some better model of rationality. Plenty of discussion and research on this in economics and philosophy.
My own take on it is that what matters in nature is not so much rationality as being adequate: as long as you do the right thing in the right situation everything is fine, and you will have many offspring carrying your genes. If you do it for completely wrong reasons it doesn't matter, as long as the error doesn't ruin things in the important areas (for example, being overly altruistic towards tribe members and xenophobic against others implements reciprocal altruism fairly well without any big demands on memory). However, this means that we come with plenty of built-in short-cuts and crude tweaks that worked really well in the Pleistocene but might be useless or worse today. We can approximate various forms of pure rationality by thinking very hard, but we are not terribly good at it.
—