SoylentNews is people
SoylentNews
khallow (3766)
First, I agree that there was widespread lawbreaking that day, some quite serious (assault, planting of the two pipebombs, felony trespass, and some novel ones coming from the penetration of the Capitol building during this ceremony). And that the people involved in the protest were for the most part, pretty stupid. But I think it's telling that only one person, an unusually stupid protester, died of violence that day.
The narratives that have been spun to portray this as something more are foundering. For example, apparently the zipties that were supposed to indicate intent to kidnap politicians were left by law enforcement. A considerable portion of the firearms and Molotovs seized that day (to demonstrate supposedly the armed insurrection) were from one person's truck (in fact, I see no evidence in the news that anyone else brought Molotov cocktails to the event). And of course, nobody actually blew up or burned anything with these explosive devices or Molotov cocktails.
The police officer who supposedly died of blunt trauma from a fire extinguisher, didn't (though police are refusing to release the actual cause of death at this time).
Much of the supposed intent is derived from chants and props that the protesters brought with them, like a gallows. Or "Hang Mike Pence!"
But those words and props never led to actions. Pence nor other politicians were hung. Nor is there any evidence that these protesters tried to capture someone for kidnapping, execution, or whatever.
Human discourse is routinely filled with exaggeration. And when you have a bunch of relatively dumb people who are angry and emotional, you'll get relatively dumb statements and imagery.
We'll see if at some point, the police or future court trials reveal evidence of insurrection or other big crimes, but until then, the media has exaggerated this story to an unusual degree. It's time to step back and take everything associated with accusations towards the protesters with a grain of salt.
For me, my eyes were opened to the subject by Hans Rosling [added link to Wikipedia entry] who over the past couple of decades before his death in 2017, popularized knowledge of the above. Unfortunately, it appears that most of his discussion is in the form of videos, but some of those have transcripts.
The TL;DR of these is that there is a global shift in almost all countries to greater wealth per person, lower fertility per woman, and a great slowing in the percentage growth rate of humanity. Supporting evidence for that is an almost linear growth rate for 60 years (which means declining percentage growth rate), said improvement in wealth per person, smaller family sizes globally, and a global shift in age distribution to older generations. Further, in the first video Rosling describes an ongoing slide in these metrics over the decades to more positive values. These trends continued after the date of that talk (in 2006). So in particular, we're not on the out-of-control exponential population growth train. We have an opportunity to get things to work out.
I'll finish with a quote from the transcript for the first talk describing the above slide.
This is where I realized that there was really a need to communicate, because the data of what's happening in the world and the child health of every country is very well aware.
So we did this software, which displays it like this. Every bubble here is a country. This country over here is China. This is India. The size of the bubble is the population, and on this axis here, I put fertility rate. Because my students, what they said when they looked upon the world, and I asked them, "What do you really think about the world?" Well, I first discovered that the textbook was Tintin, mainly.
(Laughter)
And they said, "The world is still 'we' and 'them.' And 'we' is the Western world and 'them' is the Third World." "And what do you mean with 'Western world?'" I said. "Well, that's long life and small family. And 'Third World' is short life and large family."
So this is what I could display here. I put fertility rate here -- number of children per woman: one, two, three, four, up to about eight children per woman. We have very good data since 1962, 1960, about, on the size of families in all countries. The error margin is narrow. Here, I put life expectancy at birth, from 30 years in some countries, up to about 70 years. And in 1962, there was really a group of countries here that were industrialized countries, and they had small families and long lives. And these were the developing countries. They had large families and they had relatively short lives. Now, what has happened since 1962? We want to see the change. Are the students right? It's still two types of countries? Or have these developing countries got smaller families and they live here? Or have they got longer lives and live up there?
Let's see. We start the world, eh? This is all UN statistics that have been available. Here we go. Can you see there? It's China there, moving against better health there, improving there. All the green Latin American countries are moving towards smaller families. Your yellow ones here are the Arabic countries, and they get longer life, but not larger families. The Africans are the green here. They still remain here. This is India; Indonesia is moving on pretty fast.
In the '80s here, you have Bangladesh still among the African countries. But now, Bangladesh -- it's a miracle that happens in the '80s -- the imams start to promote family planning, and they move up into that corner. And in the '90s, we have the terrible HIV epidemic that takes down the life expectancy of the African countries. And the rest of them all move up into the corner, where we have long lives and small family, and we have a completely new world.
This gets back to what I said about us gradually becoming a two class society. You have one group of people that are utilizing the countless tools this society offers to enrich themselves and they are seeing rewards like never before. And then you have another class of people who sit around, play on the internet all day, complain they don't have as much as other people, and are increasingly overtly asking governments to take it from other people and give it to them.
Looks to me like the developed world has changed. It's not so much a society of haves and have nots, but rather a society of does and does not. What remains a mystery to me is why the "do nots" think the rest will bother to support their lifestyle -not whether they should (I think I understand the moral argument here such as it is), but rather will.
For example, in the course of my work, I routinely run into people who are in their twenties and have never worked a job in their lives. Not only are they unprepared for a job, they usually are unprepared for interacting with people different from themselves (particularly, guests). The good sign is that most can and do learn. Then there are people, sometimes very advanced in age (60s, for example) who have yet to do that.
I imagine a lot of the support for UBI comes from people who just want to keep avoiding that particular thing. It probably also explains some of the people here who have so much trouble with other peoples' viewpoints.
My take is that nowadays, the most advanced societies of the world are spending considerable effort to create adult babies. We'll see how that turns out, but my small view of that doesn't look pretty.
Some of it is the weakness of proponents' arguments. Once you get past the cost savings and income inequality arguments, it devolves to things like everyone spends more time on their hobbies, which are implied to have value, while roundly ignoring that so many peoples' ideas of hobbies is watching TV on the sofa while consuming one's favorite recreational drug.
But I think there are several key problems which never get addressed. So I'll talk about those today.
The Set Point:
At its basic level, UBI is income redistribution. Someone gets taxed and then everyone gets a mostly equal share of the tax revenue. That share is something like the temperature control on an air conditioner (AC). One can raise or lower the set point to get less or more cooling. ACs have natural limits. In addition to the physic limitations of the cooling equipment, we have cost of electricity and human comfort levels. The end result typically is that the temperature ends up on the warm side of whatever is considered comfortable which also typically is the cheaper side electricity-wise.
For UBI, there's a vote disparity. It's likely that relatively few people will pay more for UBI than they receive (because rich people are relatively uncommon and there always is some degree of wealth and income inequality), so likely a large portion of voters will have a strong incentive to increase UBI. Nobody has come up with a mechanism for slowing down this demand for more UBI, and raising the set point to arbitrarily high amounts.
A second point related to this, which is more addressable, is that unlike an AC, government can borrow from the future. That leads potentially to the problem of raising UBI and then paying for it with the future.
While this sounds more serious than the former problem, it's fairly easy to solve by making UBI revenue neutral. That is, tax revenue equals UBI payout.
So my view here is that a revenue neutral UBI would still have the problem that there's no consistent way to set the point at which the UBI should be, combined with built-in incentive to arbitrarily raise UBI levels.
One possible though hard-to-implement way to solve that is a fixed tax portion. Among other things, it would give the public incentive to figure out how to improve the economy so that more tax revenue is generated rather than just raise taxes. There is a potential here for turning a UBI into a system that encourages long term thinking.
Inflation:
There are numerous factors which fall in this category. There are inflationary effects on certain goods from income redistribution. There are inflationary effects from any increase in the cost of labor (reduction in supply of labor is a likely consequnce of UBI). And a likely rationalization for raising the UBI set point would be inflation and cost of living arguments.
We already have numerous US government sectors whose cost has risen faster than the size of the economy, much less inflation, and I think it'd be instructive to consider these and how similar dynamics could cause unsustainable rises in UBI. For example, consider health care, education, and aerospace industry. The first happened due to a combination of severe supply restriction and significant increase in demand with substantial government-created cartels/monopolies, sky-high liability and testing costs, and kick-back schemes contributing to the overall costs.
Education costs rose due to massive, subsidized demand for college education. Colleges often can't even figure out what to do with the money they get - dumping it into buildings, bureaucracies, and flashy services nobody needs.
Aerospace has the peculiar problem of metric feedback. The US government has created several metrics for estimating the inflation of costs of aerospace projects. The catch is that the inflation metrics are then used to increase the funding of the projects, which in turn creates more inflation caught by the metric. It's a nasty feedback mechanism that has resulted in massively overpriced projects in recent decades.
My take is that all of this can be addressed by 1) fixing the tax rates and letting the tax base inflate normally, and 2) reversing government policies that artificially increase cost of living (I just named three). But if we were to manually try this, the various problems of built-in excess inflation come in. UBI has normally some modest inflationary effects. Combine that with policies that artificially inflate cost of living or the UBI payout, and you have a destructive feedback that could end the US in a few decades.
The Meddlers:
I've already implied some of this with the set point paragraph. Once you have UBI, you have incentive to meddle with it for profit and gain. In addition to just turning UBI to 11, there's room for carving out all kinds of restrictions: felons, legal immigrants, rich people, people who don't work, people who do, etc. This incidentally is a means for creating dissension to undermine and reverse UBI policy - deliberately create inequality in order to break the system.
The People's Skill Set:
A final thing which gets missed is the deep skill set of the working public. It goes from simple stuff like following a schedule and showing up at a place on time or knowing how to interact with other people in a public setting, to managing people or having a deep knowledge of some business sector.
My take is that a UBI, particularly a large one, is going to encourage people to stay away from work and hence, from a big opportunity to learn these skills. Then where are we going to get competent, skilled people from?
One of the things that is powerful about a democratic, capitalist society is that people are unusually competent and empowered. Even if your government is thoroughly incompetent, there's plenty of people in the private side who can do it. UBI puts that at risk by growing the portion of society that doesn't learn those skills.
What's the value of UBI?:
Too often all I hear about UBI is how it's just a right that everyone should have and how awesome it's going to be when no one has to work ever again. But what's the value to us? Particularly, to the ones who have to pay for it? My TL;DR is not that UBI is impossible to implement in a healthy way, but rather that I've heard so little from proponents about how they're going to address these huge systemic problems. My take is that in the absence of such, UBI will merely be a late stage phase in the decline of developed world countries.
There's plenty of places for "recycling" electronics, which mostly sounds like grinding stuff to powder to get the relevant materials back out.
So is there recycling of relatively high value components? For example, LED lamps (particularly of the cheap sort) often fail not due to the burning out of the LED component, but failure of other electronics (like capacitors). It sounds like one could pull a few dollars out of said LED lamp by extracting the working LED component and plugging it into a new circuit.
Speculate.
PS, *sticks lower lip out* why does JMH get his journals into search results and I don't get mine? *whine*
Way back when, Tyson Timbs was arrested and plead guilty to selling heroin (to undercover police officers). They then seized his Land Rover even though it was purchased via an inheritance rather than with drug money. He has been in court ever since to get it back.
Well, despite all that, Timbs got his Land Rover back and it only took him seven years and three different courts (with repeated bouncing around). The State of Indiana has appealed the present ruling by the lowest court to the Indiana Supreme Court which hears this case for the third time.
So why is Indiana fighting this so hard even though they are so clearly in the wrong? Because big money is at stake. My take is that they'll lose a major funding source for their police departments, if they can't steal like this. And that will have to be covered by tax revenue.
But the “extraordinary” measures we are seeing are not all that extraordinary. Powerful forces were pushing toward greater censorship and surveillance of digital networks long before the coronavirus jumped out of the wet markets in Wuhan, China, and they will continue to do so once the crisis passes. The practices that American tech platforms have undertaken during the pandemic represent not a break from prior developments, but an acceleration of them.
As surprising as it may sound, digital surveillance and speech control in the United States already show many similarities to what one finds in authoritarian states such as China. Constitutional and cultural differences mean that the private sector, rather than the federal and state governments, currently takes the lead in these practices, which further values and address threats different from those in China. But the trend toward greater surveillance and speech control here, and toward the growing involvement of government, is undeniable and likely inexorable.
In the great debate of the past two decades about freedom versus control of the network, China was largely right and the United States was largely wrong. Significant monitoring and speech control are inevitable components of a mature and flourishing internet, and governments must play a large role in these practices to ensure that the internet is compatible with a society’s norms and values.
and
Apple and Google have told critics that their partnership will end once the pandemic subsides. Facebook has said that its aggressive censorship practices will cease when the crisis does. But when COVID-19 is behind us, we will still live in a world where private firms vacuum up huge amounts of personal data and collaborate with government officials who want access to that data. We will continue to opt in to private digital surveillance because of the benefits and conveniences that result. Firms and governments will continue to use the masses of collected data for various private and social ends.
The harms from digital speech will also continue to grow, as will speech controls on these networks. And invariably, government involvement will grow. At the moment, the private sector is making most of the important decisions, though often under government pressure. But as Zuckerberg has pleaded, the firms may not be able to regulate speech legitimately without heavier government guidance and involvement. It is also unclear whether, for example, the companies can adequately contain foreign misinformation and prevent digital tampering with voting mechanisms without more government surveillance.
The First and Fourth Amendments as currently interpreted, and the American aversion to excessive government-private-sector collaboration, have stood as barriers to greater government involvement. Americans’ understanding of these laws, and the cultural norms they spawned, will be tested as the social costs of a relatively open internet multiply.
COVID-19 is a window into these future struggles. At the moment, activists are pressuring Google and Apple to build greater privacy safeguards into their contact-tracing program. Yet the legal commentator Stewart Baker has argued that the companies are being too protective—that existing privacy accommodations will produce “a design that raises far too many barriers to effectively tracking infections.” Even some ordinarily privacy-loving European governments seem to agree with the need to ease restrictions for the sake of public health, but the extent to which the platforms will accommodate these concerns remains unclear.
We are about to find out how this trade-off will be managed in the United States. The surveillance and speech-control responses to COVID-19, and the private sector’s collaboration with the government in these efforts, are a historic and very public experiment about how our constitutional culture will adjust to our digital future.
What's bizarre about the article is that the authors come up with numerous examples of censorship gone wrong and government abuse of power (the massive Chinese censorship apparatus, Snowden revelations, ubiquitous digital monitoring, and Russian government propaganda concerning US elections), and then speak vaguely of the harm of digital speech. Somehow from that, they argue that growing censorship of free speech is better than the alternative. Are they listening to themselves?
While one can argue that a private platform has a legal right to censor any way that it feels like (ignoring that for years, most of these platforms presented themselves as free speech communities and are now bait-and-switching hard), this article above illustrates one of the big ways that fall afoul of free speech law. The authors advocate for government getting involved in the censorship, both regulating it and helping the censors find the right targets to censor. They're basically calling for a Fascist-style government-lead effort to censor. That falls in First Amendment territory in the US.
And they don't seem to get that they'll probably be first against the wall when the coming tyranny needs to get rid of its formerly useful idiots.
Well, I've procrastinated quite a bit from my previous journal, yet despite that, I've managed to come across a couple of interesting, though rather obvious things. While looking for inspiration, I ran across a paper with the lengthy title, Leveraging asynchronous parallel computing to produce simple genetic programming computational models. The idea was to run a genetic algorithm (GA) on a parallel computation system completely asynchronously, by generating new programs from a pool of existing programs, and putting the more successful ones back into the pool as they are generated.
The key property is that successful programs that complete faster can return to the pool quicker (presuming of course, that you have a lot of processors/threads generating these programs rather than just one) and have the potential to evolve faster than longer running programs. So in theory, one doesn't need to include explicit time constraints in their criteria for successful programs because there's this built-in bias towards faster running programs (as long as they are successful enough to stay in the pool). The gist of the paper is that this is indeed the case in practice as theory (tested on eight different problems IIRC).
Among other things, this is in large part how real world evolution works, particularly at the microbe level. At the macroscopic level, you get some survival filters, particularly seasonal stuff, that shapes things like when organisms breed and how long they live.
From my provincial point of view, it was eye-opening because I hadn't given a great deal of thought as to how to structure the genetic algorithms that I'll attempt to start my bootstrapping scheme. The old approach that I did many years ago, the "Generational GA", was to generate a large pool of "genes", pick the best 5-10% of them, put that group in a new pool, generate from them a new pool (next generation), and start over. Now, though I'm not doing any parallelism, I can still greatly reduce the memory load by just dealing strictly with the "best" pool from the start. Generating a new gene from two in the pool, the "Incremental GA", and if the new one is better than the worst gene, replace the worst. In addition, it appears to be a little faster in converging to a useful solution (though I haven't quantified it).
The Lattice Coloring Problem
Finally, let me briefly describe what I'm presently working on. I start with a 60 by 60 grid (the number 60 is chosen because it allows for patterns with periods up to 6) of "colorings", every vertex is numbered between 0 and N-1 (I presently use N=4). A small patch of the grid looks something like:
0 3 2 3 1
1 3 3 1 1
2 0 0 3 1
3 0 2 2 1
The energy of this lattice is based on a randomized function that takes four adjacent vertices of a 2 by 2 box and assigns a real number to them.
w x
y z
Then sum over all such boxes of the grid, wrapping around both top and bottom, left and right. For the 60 by 60 grid, you end up doing 3600 such sums.
The problem then is to find a color pattern that minimizes this total energy. One gets genuine random colorings as the lowest energy state only for rare cases (such as when all possible colorings have the same energy). Instead interesting patterns appear.
Some common examples (just providing the pattern, not the energy function that these patterns happen to be minimum energy for):
1 1 1 1
1 1 1 1
1 1 1 1
1 1 1 1
Solid color is the most common.
1 2 1 2
1 2 1 2
1 2 1 2
1 2 1 2
Alternating solid rows or solid columns is next.
0 3 0 3
3 0 3 0
0 3 0 3
3 0 3 0
Checkerboard.
2 3 2 2 2 2
2 2 2 3 2 3
2 3 2 2 2 2
2 2 2 3 2 3
A pattern with a bit of randomness thrown in.
On the last one, the building block is a 2 by 2 block of 2s with one 3. You can shift either the rows or the columns (but not both), randomly offset by 1 (here, it's columns), and still have a minimum energy. I've looked a little bit for a genuine nonperiodic pattern (which would be a quasicrystal), but haven't found it yet.
Here's an example of a four color coloring that can be a lowest energy state:
0 0 0 0
1 1 1 1
2 2 2 2
3 3 3 3
The nice thing about this particular problem is that it's very easy to turn into a genetic algorithm. I basically just copy/paste random rectangles from one grid to another. I can also mutate by copy/pasting from a randomly generated grid to a good grid. It's one of the simpler problems one can apply the GA approach to.
So how fast does it converge? Not very. But starting with a pool of ten randomly generated grids, I usually can come up with a reasonable try after a few hundred to few thousand iterations. What I've discovered is that convergence is fastest when the lowest energy pattern isn't near other patterns in terms of energy. For example, if all 2s is almost as low an energy as all 1s, we would see the 0 and 3 colors rapidly disappear, but then it'd take longer for the 2s to vanish. Sometimes the program would halt on all 2s or even not get to a state of all 1s or all 2s.
I've also had difficulty reaching complex pattern lowest energy states. They tend to converge slower than for the simpler color patterns mentioned above. And if the pattern has a period that doesn't divide evenly into 60 (such as 7 or 9), the pattern will always be truncated. That can introduce higher energy defects which take a while to settle down.
As a final remark, I could look at different numbers of colors or different sizes of grid, as well as figuring out how fast these things converge under different circumstances. I haven't bothered at this time to do so.
What's Next?
Presently, I'm still avoiding actual genetic programs and am instead looking at fitting a table of data (each column corresponding to a parameter) to a polynomial function of the parameters. It's still a pretty easy problem for GA to attack, but involves generation of arbitrary monomials which involves similar tree-building to what generating programs would involve. I presently am thinking of using it to generate a differential equation for a vector differential equation I messed with for a while. Mathematica doesn't seem effective at dealing with it in its present form. Maybe, if I convert the problem to a bunch of separate ODEs, it'll work better.
I'm also thinking about making a default mutation algorithm of breeding a gene with a randomly generated gene since I need both subalgorithms anyway for my basic GA system.
Well, I can't do a good review of what the current state of AI is. However, we can look at some sample cases. For example, the much-hyped IBM Watson system is initially trained by taking a huge database of human information, and then training it to come up with answers (its interface is a natural language question/answer system).
Once it's fed and running, it allegedly works quite well at the niche problems it's tasked to solve.
If we are to consider the bottlenecks in this process, a key one is the design of Watson in the first place. Namely, it takes considerable effort of a vast number of programmers to put together the system. Any significant improvements in the system probably will require that same army. That is, if you want to make Watson better, you need man-power.
This is common to a number of the current approaches (neural nets being another example). You have to work a lot to get the system to the point where you can feed it data. And if you want to make it better? You need to do the work yourself.
So how to fix that? An obvious way, around for decades, is to throw the power of the software at improving itself, or bootstrapping. There are several closely related definitions of the word. A common one is an installer that downloads and activates a larger program that does the actual installation. Here, I'll treat said software as a more or less generic partial optimizer with the ability to apply itself to itself iteratively at the speed of the computer not of the programmer. That's the bootstrap.
So I started thinking how could a small group or even a single person implement a bootstrappable system? So here's my high level scheme:
- Create or use an existing homoiconic language. By definition, these are languages where code is readily manipulated as any other sort of data by the base language itself. I was looking at using an modestly extended version of Curry combinators (with stochastic and I/O combinators added) for simplicity. Programs could then be thrown as data to programs of the same kind.
- Construct a rudimentary optimizer for the initial kick. I'm feeling the genetic algorithms approach here, but if completely desperate could randomly generate programs and hope something sticks to the wall before the heat death of the universe.
- Find or make a standard structure to describe optimization problems and construct a useful random optimization problem constructor.
- Go meta. Using the best optimizer I have, use it to construct a population of optimizers that then compete on the random optimization problems to determine the next generation of best optimizer (I'm expecting a population rather than a single one because I'm using a stochastic process which has more than one possible output). I can similarly optimize my random optimization problem constructor in the same fashion to come up with more challenging, useful, or complex optimization problems.
- Figure out what's going on.
The idea is to take the human out of the loop near completely and see what happens. Since I think the basic building block, the Curry combinator is rather inefficient, the point is more to build a model of how bootstrapping works than to build useful, efficient code (my primary contribution to efficiency is to pick a system that isolates the resource-consuming copy combinator) - though I have no problems with that, if it should be a consequence.
A key problem is that we don't know what to expect from bootstrapping, particularly metrics for things like rate of improvement. Some samples, even if they aren't practical, would help us come up with conceptual models of bootstrapping. For a crude example, suppose we can measure the rate at which the productivity of our efforts increases as a function of our present productivity. If that function slows down enough, we'll cap at a near future amount little better than present. If it's linear, that's exponential growth (at least as long as it lasts). If it's greater than linear (at least for a stretch), then there's an genuine near-singularity in the near future.
Some useful features down the road would be ways to log information about generations (I'm sticking with the genetic algorithms paradigm though there's no reason to expect future generations of optimizers to strictly follow it).
If there's interest, I can go on. At this point, my hope is that this is close to simple enough that people will look at it and think "Hey, I can do that" and maybe we can bust open this particular concept (and how to measure it) in a way that doesn't require an army of coders.
