Captain_Maclaine posted:

I don't think he's convincingly shown there to be one, no. It's why I called his argument a solution in search of a problem, honestly.

quote:

Recasting all complex social situations either as neat problems with definite, computable solutions or as transparent and self-evident processes that can be easily optimized--if only the right algorithms are in place!--this quest is likely to have unexpected consequences that could eventually cause more damage than the problems they seek to address. I call the ideology that legitimizes and sanctions such aspirations "solutionism." I borrow this unabashedly pejorative term from the world of architecture and urban planning, where it has come to refer to an unhealthy preoccupation with sexy, monumental, and narrow-minded solutions--the kind of stuff that wows audiences at TED Conferences--to problems that are extremely complex, fluid, and contentious. These are the kinds of problems that, on careful examination, do not have to be defined in the singular and all-encompassing ways that "solutionists" have defined them; what's contentious, then, is not their proposed solution but their very definition of the problem itself. Design theorist Michael Dobbins has it right: solutionism presumes rather than investigates the problems that it is trying to solve, reaching "for the answer before the questions have been fully asked." How problems are composed matters every bit as much as how problems are.

evilweasel posted:

that would just ruin the fun

do you ask if there is a basic need to stop the football mans from putting the ball in the painted section of the grass?

Cream_Filling posted:

I'm buying this book now, actually, but:

quote:

Recasting all complex social situations either as neat problems with definite, computable solutions or as transparent and self-evident processes that can be easily optimized--if only the right algorithms are in place!--this quest is likely to have unexpected consequences that could eventually cause more damage than the problems they seek to address. I call the ideology that legitimizes and sanctions such aspirations "solutionism." I borrow this unabashedly pejorative term from the world of architecture and urban planning, where it has come to refer to an unhealthy preoccupation with sexy, monumental, and narrow-minded solutions--the kind of stuff that wows audiences at TED Conferences--to problems that are extremely complex, fluid, and contentious. These are the kinds of problems that, on careful examination, do not have to be defined in the singular and all-encompassing ways that "solutionists" have defined them; what's contentious, then, is not their proposed solution but their very definition of the problem itself. Design theorist Michael Dobbins has it right: solutionism presumes rather than investigates the problems that it is trying to solve, reaching "for the answer before the questions have been fully asked." How problems are composed matters every bit as much as how problems are.

Tokamak posted:

You can use a computer to solve what a group of people should eat at a restaurant, which makes everyone happy. There are perhaps 20 people, and 50 menu items and each person has a few preferences and one restriction.

RealityApologist posted:

RealityApologist posted:

In this thread in particular, I've not suggested any magical computer dust that fixes all problems, or dismissed any of the potential problems being raised, or made any of the nonsensical claims being proposed. If people are getting that impression from the thread, it's entirely the responsibility of the obfuscatory insults being hurled my direction. It's funny to read you guys tossing insults back and forth, and then pausing to say "you know, this would be interesting if Eripsa wasn't so crazy". If I leave y'all alone, you might actually have a productive conversation about this stuff.

In particular, I agree almost entirely with 1337JiveTurkey's post, and I'm curious to know why he thinks I've said anything that contradicts it. He says basically exactly what I've been trying to say, albeit more clearly written, but then he says "but obviously I disagree with Eripsa" as if I've said anything different. So I suppose I'll just leave this thread alone.


I used the big red Norvig book in my AI class a decade ago, and it still sits on my self. I've even pulled it out once or twice since. I feel I'm pretty familiar with the problems of computational complexity and optimization. I'm not an expert, of course, and I'm happy to learn anything contributed to the discussion, but I don't think I'm over my head for what's required for a thread in SA.

Anyway, you are right, this is a computationally complex problem, especially if you assume that some fixed meal (or some fixed range of meals) will make each individual happy. It gets even more complex if you think that "what makes a person happy" is itself a function of the meals taken by other people in the group, but that seems to be a realistic extension of the problem: Jimmy doesn't just like potatoes in the abstract, but will get potatoes only if Suzie gets potatoes because he's trying to impress her, or whatever. This extension of the dinner problem, where individual choices are responsive to the choices of all other agents, increases the complexity of the task exponentially. So yeah, these are difficult problems.

But just last week, families got together all over this country, often with 20 or more people, and potentially with dozens of thanksgiving dinner items to select from. And in fact, they are all responding to an enormously complex number of subtle relations and constraints that are far densely layered than anything close to the abstract dinner problem we described above. And these people were all tasked with the problem of sorting out the food to meet preferences and make everyone happy. And they solved the problem. Easily. Many while raging drunk.

It's important that human beings eat together in these kinds of social rituals precisely because it gives our brains a chance to engage in exactly these kinds of complex social networks. That's not only a large part of what our brains were built to do, but it also established the real social ties that support my real daily life. It's because I've engaged with these rituals with friends and family that I have the social support network I do. Human computation is about incorporating human beings into the computational system, because some of these outrageously complex computational tasks can be solved quickly and easily solved by collections of human brains working together. A network where people are self-motivated to participate tends to be healthy and supportive of its members, and tends to reduce the alienation and isolation that is typical of more hierarchical modes of control. We need human computational systems because, counter intuitively, it makes us human again.

So here's a solution to the dinner problem: take 20 hungry people, inform them of their preferences and restrictions, put them in front of 50 plates of food, and let them go at it. I guarantee you this process won't end in an infinite loop. Although the computational complexity of the task might be exponentially large, the actual solution time in all empirical trials of the above algorithm average to constant time, with surprisingly low c.

This is why self-organization matters as part of the theoretical structure of the view. An attention economy operates on the principle that the human behaviors themselves are doing an enormous amount of the computational work required, especially about selection criteria and determining what's "important". Through their behavior, they are indicating what they consider worth their time, and why resources they are willing to devote to that issue. The problem for designing a system is not how to compute these results ourselves, but rather how we can harness the work already being done by all the people, and leverage it for organizing the whole system.

SedanChair posted:

RealityApologist posted:

Cream_Filling posted:

I'm buying this book now, actually, but:

RealityApologist posted:

Morozov is right to point out the political darkside to a lot of the technology that is overlooked by the optimistic bullshit, but his critique is not supplemented at all by a constructive story about how it might be possible to use these tools to improve our situation. It's all sneer, and no constructive anything. At the end of the day if that's all he has, then its just a style thing, so that he can be that talking head when the media needs someone to play the role of the skeptic.

Well, gently caress that. We have problems that need fixing, and we're gonna have to use whatever we have handy, and we just happen to have a lot of really powerful computers in our hands most of the time. Let's really start to think about what we can do with these things. Facebook and twitter are to the early digital age what hot air baloons were for early flight. Cool that we can do it, lots of engineering challenges and all that, but come on, its not really flying. You don't have to be a singularity mystic to think we can do significantly better. Morozov won't do anything to help with that project because he's positioned himself so that it would be a poor career choice. So gently caress him too.

RealityApologist posted:

I've not proposed any particular solutions. I've proposed procedures by which solutions might be found. This is not a solution itself; it is a metasolution, the form of a solution.

RealityApologist posted:

I've not proposed any particular solutions. I've proposed procedures by which solutions might be found. This is not a solution itself; it is a metasolution, the form of a solution.

Cream_Filling posted:

I like the part where he claims that not having any details actually shows his complexity and nuance because he left those bits blank to be filled in as appropriate to the situation.

Cream_Filling posted:

*throws a bolt of cloth on the table*

"Here's your dress. Now sit down and I'll describe what a dress looks like so you can finish it. It's kind of a big dangly thing that covers up your body-- what, you were expecting some sort of one-size-fits-all solution? It's a community-oriented approach."

Cream_Filling posted:

Uh humans don't do a lot of actual computation they mostly rely on heuristics and cultural customs. No person sits down and calculates their optimum menu based on relative preference rankings, not even ulillia. They mostly just pick whatever comes to mind that's culturally appropriate and easily available. It's not particularly accurate but it's fast and good enough. Sit a bunch of people down in front of plates and tell them to pick one each and some will pick whatever's closest, some will pick the shiniest or brightest colored one, and some will look at their neighbors to decide. Most will probably not be acting optimally or rationally, but this isn't a big deal.

The obsession with efficiency means people look for ways to find optimum choices using computers, but this is a massive computational problem. Humanity gets around this (in part) by not actually doing all the computations and sacrificing perfection for speed. Except when a computer tries to do this, often all that happens is you get an inferior or at best almost-as-good-as-human result because humans have been doing this for a long time and also because we're far more tolerant of (and able to compensate for) the kidn of mistakes humans make versus the ones computers make. With enough work maybe you can consistently top human decision-making in some narrowly defined field where there's known weaknesses in human thinking and/or lots of ways to obtain feedback and additional decision data, but even then human preferences are tricky bastards and constantly changing so overall it's quite ambitious to set out to do so, though not necessarily impossible. But to do so for all things everywhere? Just plain absurd.

Also, ritual is actually sort of the opposite of doing additional computation or decision-making, too, since rituals are distinguished by actions that have a prescribed order or form, usually functioning to restrict and direct action and thought.

DrProsek posted:

It'd be a pretty awesome project to work on not gonna lie.

RealityApologist posted:

Now maybe it's absurd to think that we can use this kind of system to manage all the things, but my muddled brain and my smartphone together seem capable of contributing some powerful computational work, if harnessed correctly. I mean after all, we did use these tools to rate and review a huge chunk of the products and services available to us over the last few years, and we've taken pictures of quite a nauseating amount of things. At the rate we're producing data, it's hard to imagine anything else.

RealityApologist posted:

But just last week, families got together all over this country, often with 20 or more people, and potentially with dozens of thanksgiving dinner items to select from. And in fact, they are all responding to an enormously complex number of subtle relations and constraints that are far densely layered than anything close to the abstract dinner problem we described above. And these people were all tasked with the problem of sorting out the food to meet preferences and make everyone happy. And they solved the problem. Easily. Many while raging drunk.

1337JiveTurkey posted:

The United States ITYOOL 2013 is not a compelling example of congestion usually being a result of the failure of top-down rules. In spite of all the top-down efforts taken across the country, there's still highly periodic bouts of congestion in nearly every major city. Rush hour and certain holiday weekends happen without any central direction telling people what roads to drive on. The same patterns occur in spite of policy differences between cities, and sustained changes within a city still eventually stabilize in a new pattern while transient disruptions tend to revert to the norm. It's self organized in the sense that there's no centralized top-down system telling everyone which roads they must drive on for a particular trip and it's more robust than the centralized system crashing and not letting anyone drive anywhere.

Having said that, I'm not agreeing with Eripsa. It's robustly lovely and sprinkling magical computer dust on everything will only solve the problem insofar as the swarms of omniscient mind-reading murderbots reduce the number of people driving in the first place.

DrProsek posted:

It'd be a pretty awesome project to work on not gonna lie. Kind of in the same way Bitcoin is a cool project: it would be sweet as hell to get to work on the project, and turn this wacky theory into a working proof of concept just to say "son of a bitch, this poo poo actually works!". The problem is when someone runs with it and says the world economy should work off of it that all the logistical problems of running the world's attention data through a central network poke through and show you all the ways this (for the forseeable future) cannot and should not be used wide scale.

RealityApologist posted:

Anyway, you are right, this is a computationally complex problem, especially if you assume that some fixed meal (or some fixed range of meals) will make each individual happy. It gets even more complex if you think that "what makes a person happy" is itself a function of the meals taken by other people in the group, but that seems to be a realistic extension of the problem: Jimmy doesn't just like potatoes in the abstract, but will get potatoes only if Suzie gets potatoes because he's trying to impress her, or whatever. This extension of the dinner problem, where individual choices are responsive to the choices of all other agents, increases the complexity of the task exponentially. So yeah, these are difficult problems.

But just last week, families got together all over this country, often with 20 or more people, and potentially with dozens of thanksgiving dinner items to select from. And in fact, they are all responding to an enormously complex number of subtle relations and constraints that are far densely layered than anything close to the abstract dinner problem we described above. And these people were all tasked with the problem of sorting out the food to meet preferences and make everyone happy. And they solved the problem. Easily. Many while raging drunk.

It's important that human beings eat together in these kinds of social rituals precisely because it gives our brains a chance to engage in exactly these kinds of complex social networks. That's not only a large part of what our brains were built to do, but it also established the real social ties that support my real daily life. It's because I've engaged with these rituals with friends and family that I have the social support network I do. Human computation is about incorporating human beings into the computational system, because some of these outrageously complex computational tasks can be solved quickly and easily solved by collections of human brains working together. A network where people are self-motivated to participate tends to be healthy and supportive of its members, and tends to reduce the alienation and isolation that is typical of more hierarchical modes of control. We need human computational systems because, counter intuitively, it makes us human again.

So here's a solution to the dinner problem: take 20 hungry people, inform them of their preferences and restrictions, put them in front of 50 plates of food, and let them go at it. I guarantee you this process won't end in an infinite loop. Although the computational complexity of the task might be exponentially large, the actual solution time in all empirical trials of the above algorithm average to constant time, with surprisingly low c.

This is why self-organization matters as part of the theoretical structure of the view. An attention economy operates on the principle that the human behaviors themselves are doing an enormous amount of the computational work required, especially about selection criteria and determining what's "important". Through their behavior, they are indicating what they consider worth their time, and why resources they are willing to devote to that issue. The problem for designing a system is not how to compute these results ourselves, but rather how we can harness the work already being done by all the people, and leverage it for organizing the whole system.

Main Paineframe posted:

This is largely a function of insufficient capacity, though - it means that either the major roads weren't designed to handle enough cars to prevent congestion during rush times, or some unexpected event (such as a car crash) is preventing the road from handling its full capacity. The cause for this usually has nothing at all to do with driver behavior; the former is usually due to decisions made during the top-down planning process, while the latter is better attributed to the "Acts of God" clause rather than driver self-organization.

Main Paineframe posted:

This is largely a function of insufficient capacity, though - it means that either the major roads weren't designed to handle enough cars to prevent congestion during rush times, or some unexpected event (such as a car crash) is preventing the road from handling its full capacity. The cause for this usually has nothing at all to do with driver behavior; the former is usually due to decisions made during the top-down planning process, while the latter is better attributed to the "Acts of God" clause rather than driver self-organization.

SedanChair posted:

RealityApologist posted:

T-1000 posted:

There are ways to avoid this such as having a larger buffer zone between cars to allow smoother deceleration and acceleration without a complete stop, and communication between all cars to prevent excess braking, but neither is an option presently.

edit: you could also build roads everywhere with sufficient capacity to handle the maximum number of cars they would ever have without ever reaching the critical traffic density, but that would be incredibly wasteful.

RealityApologist posted:

I understand it well enough to know the example was both reasonable and correctly illustrated the point being made. You all know it too. In a conversation, some ground must be accepted by all parties in order to make any sort of progress, but I'm not allowed to make even obvious points by way of example in order to illustrate my claims.

If my argument was attempting to make a deep point about military tactics then maybe I should be called on it. But the example was only making the very simple point that this:



is different from this:



and that's a distinction I'm perfectly qualified to observe as a layman.

FilthyImp posted:

I remember reading someone's weird online site that basically said this. He just observed how hosed up traffic got when people were bumper-to-bumper tailgaiting and noticed that if you leave a large enough gap to allow for zippy idiots and slowdowns, it all works down.

As to the second post, in Los Angeles at least, lots of that gridlock happens because there are certain main-roads that you have to resort to if you're getting anywhere. I'm not sure how the datamasters would handle a fix for that...

T-1000 posted:

Traffic jams are a function of traffic density, without the need for any sort of accident; any road can handle a certain critical density of vehicles before before james will happen. Katsuhiri Nishinaro has done some incredibly interesting work on this, he calls it "jamology". The important graph is on page 3 of the PDF I've linked, where the flow suddenly crashes after reaching a certain density of traffic. In practice, all you need is for one car to slow down, the car behind it slows down slightly more, the car behind it slightly more, and then a slight lag in each car accelerating back up to speed. There are ways to avoid this such as having a larger buffer zone between cars to allow smoother deceleration and acceleration without a complete stop, and communication between all cars to prevent excess braking, but neither is an option presently.
edit: you could also build roads everywhere with sufficient capacity to handle the maximum number of cars they would ever have without ever reaching the critical traffic density, but that would be incredibly wasteful.

Popular Thug Drink posted:

You have no idea what you're talking about and I feel pity for your students. I know this because I can see the example you're trying to make, the example you're actually making, and the wide gap between both points that you seem unable to percieve. You are not an effective communicator and if anything these poor communication skills lend you false credibility to people who cannot parse your bad arguments and simply assume your authority. You should not be talking to young adults about philosophy.

RealityApologist posted:

My point is that we don't need to solve these problems in the abstract case, or to perfect generality; we just need to solve them well enough to meet the human constraints on the system. If there's food around, people will distribute it, and other things equal they tend to distribute it in fair and reasonable ways, especially when they are very young. But yeah, the ritual where Dad carves the bird onto plates might do a lot of the deciding work for how the food gets distributed, and takes the load off everyone so they don't have to worry about at least some branches of the decision tree. There are positives and negatives to that (of course), but these are exactly the kinds of things we do to resolve these kinds of tasks.

The ritual helps perform the computation by embedding it deeper in the infrastructure so it doesn't require active attention to maintain. James makes the point about habits more generally: if I develop a habit for doing X, then I don't need to waste the mental resources of deliberately and consciously deciding to do X so that after a while it becomes a lot easier to do X. We might talk about this in terms of "externalizing" cognition, so that external resources (including things like "muscle memory") take up some of the processing role for us. The point is that our cognitive systems become increasingly entangled with these external aids to processing and memory through the habits and customs we build up around them.

Now maybe it's absurd to think that we can use this kind of system to manage all the things, but my muddled brain and my smartphone together seem capable of contributing some powerful computational work, if harnessed correctly. I mean after all, we did use these tools to rate and review a huge chunk of the products and services available to us over the last few years, and we've taken pictures of quite a nauseating amount of things. At the rate we're producing data, it's hard to imagine anything else.

Cream_Filling posted:

Also people left to their own devices don't always distribute things in a fair and reasonable way. This is a ridiculous assumption that basically manages to handwave away the actually difficult part of a problem - people left to their own devices usually do ok in small family groups but this organization totally breaks down at larger scales or without direct personal interaction and knowledge.

Cream_Filling posted:

Also if this was true even in families, "YOU WERE ALWAYS MOM'S FAVORITE" wouldn't be a thing people yell and we wouldn't need tons of lawyers and insanely complex laws for how to handle estates after people die.

Obdicut posted:

My had has a severely hosed-up relationship with food due to growing up in a depression-era family where he and his brothers fought continually over food. It's not just the technology that Eripsa gets so bastardly wrong, it's also basic human nature.

I've just been writing an essay on the ethics of privacy, so I was reminded of this thread. The 'attention economy' would innately punish people who valued privacy: this could be considered a feature if you're a nosy parker. Personally, I think that it would be yet another way that the 'attention economy' would reinforce dominant hegemony.

N. Senada posted:

I thought you said you were going to stay away from this thread.

I guess you just couldn't stop paying attention.

Obdicut posted:

In actual content, I'm rereading Darwin's Dangerous Idea, and I think Dennett's call there for a philosophy of engineering is more apt than ever, because lacking it there tends to be this sort of cranky thinking that software is magical, that organization is all around us, ready to leap out if only super-duper technology/social system X is adopted. In addition, the way he lays out the contingent nature of evolutionary systems is a strong argument that self-organized systems don't seek the most optimal solution anyway; the path of organization depends on decisions made earlier in the evolutionary path.

RealityApologist posted:

I'm a huge fan of Dennett, and it's obviously inspired a lot of my work.

My claim has never been that "if we only make our political systems self-organized, it will always find the optimal solution!". That's only ever been a poor characterization of the view, and I've been objecting to this strawman for now 13 pages, and I wish you all would stop barking it like jackals. My argument has always been about leveraging the existing organizational dynamics in populations in favor of generating the solutions we want.

quote:

Nevertheless, the science is clear enough that we can start taking seriously the questions and issues being proposed in this thread, and seriously start thinking about human computational systems and what real organizational work we can get them to do.

quote:

Again, people solve the "what's important? what's meaningful?" question continuously as they go about their lives, and with enough data points we will start constructing full scale models of the big-v Value of all the things, for all the people.

quote:

It will give us more data from which to make our decisions, more channels through which feedback might reverberate through the network, and more opportunities to specialize and address all the problems at all the scales that effect humanity.