Daily Reading #5A

Controlling risk is the key to long-term success. Black swans control your future, we are still living through the effects of WWI, the flu pandemic that followed, and WWII, … None predicted, and our near-future Black Swans will screw things up just as thoroughly, our grand children will endure their TBD effects throughout their lives. I can be confident of that prediction because it is always true, the nature of nature and complex systems. Trading is only different because it is a simple enough domain that it can have effective risk management tools. The rest of life, not so much, and thus I guarantee we humans don’t put enough mental effort into identifying the important risks and working to produce hedges. This is a good reminder of how to think about risk, he recommends ‘humbly’ above all :


One of the long-term dangers is the world of secret organizations, e.g. the CIA. This is Col. L. Fletcher Prouty’s take on his time with the CIA. He isn’t a fan, was high enough and deep enough to see the flaws in 3D. I haven’t finished it, it is a 400-page book, but it includes very much history, many examples, and all throughout keeps providing an overview. A must-read for anyone who wants to know how we got here, how to fix it. This lets you download the .pdf :


That, of course, is the post-WWII version of what 2-time Medal of Honor winner Smedley D. Butler, who wrote “War Is A Racket” recounting the economic interests that drove his wars, and so directly :


That is consistent with this more-recent critique from John Kiriakou, a more recent whistleblower within the CIA, which I linked to earlier. A major difference is that Prouty didn’t buy the base story of the CIA doing important work. Kiriakou does. Prouty would have noticed 9-11, neither Sheer, the interviewer, nor Kiriakou do, so this is, at best, a limited hangout, a small disclosure directing attention away from the more important story, the main point of prosecuting whistleblowers for espionage, the need to stop insider leaks of information on 9-11, Sandy Hook, ISIS, … so many false flags and fake News the CIA produces or directs :


Eric Zeusse, a serious Progressive, critiques Obama as a failed President. I agree, at least this bad :


This gets at least a lot of the story right.  I don’t know how Rahm Emmanuel fits into it, but he was the first of Clinton’s people at the DNC, and he started selecting pro-Israeli, neocon-thinking Democratic candidates based on his access to AIPAC donors. They mostly lost. When voters start rejecting candidates because of their donors, we might get somewhere.

The other big point in this is, the situation of money controlling politics is inevitable so long as choosing a representative is not personal, no more than 2 social links from people you know well enough to trust you to the representative.  That means no more than 25K people / representative in any legislature: : you know 30 people well enough to trust them, they each know 30 and each of them know another 30, so 27K people at 3 levels.  That allows every representative to have a local reputation, he isn’t a remote figure, and was the case in the early Republic.

There were about 210M eligible voters in the US in 2000. 25K voters max would mean 8400 Representatives in Congress. that is about 20X the current number, so the Senate should be expanded proportionately to 2000 members. That means 40 senators per State. Senators are state-wide races, so voters could chose the top 40 candidates, or states could be made smaller and the senators divided proportionately. That expansion would put power back into the hands of the local community, even senators would not need to run huge campaigns, and individual votes would not be as valuable so there would be both less need and less incentive to finance them. That reform alone would be a major way of decentralizing DC.

We need a unitary executive, and the maximum size of a working group running on consensus is probably 8, I propose selecting from a list of Presidential candidates, with the 8 who get the most votes being joint occupants of the office of the President, with the rule that all must agree on every Presidential action. That doesn’t guarantee optimal decisions, but there are literally 1000s of studies showing that groups make better decisions than any individual among them, and training can make group decision making very good indeed.

For every elected representative or executive, we need much easier recall for any possible violation of the Constitution or law.

Add those and another few changes to the Constitution, we might be able to get our government back under control :


This is a conference organized by Aubrey de Grey on prolonging lifespan. Interesting research, and the first time I have seen that much emphasis on reproducibility, quite a change from previous presentations I have watched.

The discussions and points of view were interesting, tho working through the details of how to organize cell types into tissue could make you think the process of science is detailed boredom. Compared to physical sciences and the engineering disciplines that depend on them, it is developing the handbooks of materials and tables of constants and mathematical values that engineers use. Has to be done, and the Army is giving Wake Forest a lot of $ to run that institute. From the other presentations, you mostly learn that there are evolutionary tradeoffs in every dimension, and lifespan is just another dimension in the many-D space that is being optimized by evolution’s actions, thus it is a complex problem to understand, those take a while and initial error bars are large because nobody knows what the proper controls are before they are found.

I still think that mitochondrial transfusions are a good possibility for general life-extension, although one of these papers mentions that too many mitochondria can induce apoptosis, also said to be the effect of DCA on cancer tissue :


How to become a super-ager :


Exercise will keep everyone’s VO2-Max high, necessary for a well-oxygenated brain, necessary for it to work well, also good for mood :

This is Dr. Nick Bostrom, University of Oxford, Director, Future of Humanity Institute.

His talk is very theoretical, and he doesn’t discuss the widely different arenas and differing successes.

It seems to me there are two different classes of problems, ones that can be dealt with entirely by simulation, e.g. the neural nets for Deep Learning with images, sounds and games. OTOH,  there are robots doing the dishes, making a hamburger, … many tasks, and many variations in very many dimensions in the circumstances of making a hamburger or doing the dishes. Those can’t be simulated so easily (although Google just released 3D rendering that might help that), and therefore training cost is a big issue.

Google DeepMind’s new simulated world could lead to advances in robotics, and help us better understand how humans learn.
Google DeepMind, a subsidiary of Alphabet that’s focused on making fundamental progress toward general artificial intelligence, is releasing a new 3-D virtual world today, making it available for other researchers to experiment with and modify however they wish.

Doing both the cognitive and NN levels simultaneously is difficult, analogous to learning an athletic skill, coaching makes a big difference and there is different effective coaching for the mental and physical part of the task. Ditto for purely cognitive tasks.

Do we have a theory of any of these things? No, certainly not. So expecting to develop an artificial super-intelligence without any of the theories needed to even much improve human’s effectiveness and efficiency, learning more better per unit time, more usefully for specific tasks and more generalizable, will be an evolutionary search in an combinatorially unbelieveably humongous-enormous state space.

We know what controls progress in those:

generation time  = speed of growing a new AI to a level we learn as much as is needed from it, then make the changes needed for the next experiment

the population size = number of combination trials per year

rate of mutations = incorporation of variation into the combinations.

Probably also some magnitude of mutation, but that is hard to measure, sort of how big a wild card might be, both positive and negative effects on ‘fitness’, the rate of learning by this AI for a variety of tasks would be one of many such measures.  Negative is usually larger than positive, of course.

The generation time for versions of AIs and their evaluation is at least 10 years. The population of versions of AIs under evaluation at a given time is perhaps proportional to the number of Ph.D. AI researchers, no more than a few 10s of thousands and that is probably a big over-estimate, but I can’t find a better number. The rate of mutations is certainly less than the 100-200 per new individual humans, human minds can’t make sense of changing that many things at once.  Assume it is 1/10th of the human rate, perhaps the same as the proportion of mutations affecting the brain, I will even give them the assumption that, because these are still research projects and nobody does exactly the same thing every generation, even when working on a new problem with current-gen tools, that the changes will all be positive, but I doubt that is true. Doesn’t matter, the combinatorics is the problem.

It took 6M years for intelligence to improve from chimp-level to modern humans, say an IQ of 40 (and I have seen estimates as high as 85) to 100,a 6 sigma to 1 sigma range. Assume a super-human AI is only 145 on average, 3 sigmas, so would only take half as many generations.  3M years at 20 years / generation == 150K generations. Average population was perhaps 1M individuals, and rate of mutation was 100-200 mutations per new individual out of 20K genes and perhaps 2X that regulatory regions.

1M individuals per generations * 20 brain-mutations per individuals * 150K generations = 20M * 150K = 3T tests of combinations of 20 new combinations of brains in natural evolution.

AI’s evolution, by comparison, averaging for the next 100 years of greatly expanded research to 100K individuals per generation * 20 changes per new AI * 10 generations in the next 100 years = 20M tests of combinations of 20 new combinations of AI techniques and theories in this AI evolution.

3T/20M = 150,000 multiple advantage to natural evolution, 5 orders of magnitude.

The researchers will claim ‘some theory’ and ‘directed evolution is much more efficient’. Maybe, but it is all claim, so far as I see just now.

Anyway, this guy didn’t make me think AIs are going to be grading their own papers any time soon, I don’t have to rewrite this yet :


Warren Pollack predicts Trump has inherited the Titanic, a very fragile economy and anything done to fix it will kill it. That fixing it is impossible until real prices and plain language allows planning and thinking. Also that Clinton and Trump must fight to the death, for the reason that he doesn’t kill the leader of the opposition, otherwise the left won’t stop their opposition. I agree with the prediction, but my reasoning says that Trump made a promise to ‘drain the swamp’, it is a very popular position with his voters, and I think people are going to hold him to it. Along with that will be massive opposition from all directions. The way to keep his political momentum is to have a vigorous prosecution of some key players, beginning with Hillary Clinton and a pedophilia ring or two. Republicans will be caught, enough to make it seem even-handed, but the Justice Department doesn’t have enough prosecutors to pursue everyone, just the worst, we hope.

So big loud investigations and prosecutions, modern show trials. They will have just as much system-level effect as ever, damn little at most. This is not the first time in living memory that every element of the bureaucracy has been reformed, they are well practiced at dealing with reformers. Likewise, there is never a surplus of investigators on these very interconnected cases, so many leads won’t be followed, and again, the administration’s friends will often get a pass.

If Trump wanted to have a larger effect, really clean all of the swamps all at once, he would tell the FBI to open their files and let citizen researchers begin following leads and local prosecutors and private citizens bring cases. Did you know that you can join with others to file private RICO suits against corporations and other entities? And win damages, big money damages? Another good innovation would be to remove immunities of government entities and personnel. If we are liable for breaking laws, they should be equally liable, and I can provide a good argument that they should be held to even stricter standards, it is a necessary condition for system stability over the long run.

We have a considerable surplus of lawyers in this country, it is a cheap resource that we should be using in solving our major problems. With open databases from the FBI’s research as a beginning, citizen researchers pursuing leads in public sources, and then depositions, discovery and thousands of law suits against oligarchs, governments, banks, … Combine that with open prosecution files, so that groups could share the legal work, specialized state to state, and that would prevent the major law firms defending the entities from developing more expertise than the very many distributed legal teams bringing suit against the. Optimistically, we would have things fixed in a generation or two, but it is going to take a major revamping of legal concepts in our society. We now specify process, not the requirements to be met for Justice.

There is no way the FBI + the Justice Department will even dent the corruption in our society, show trials will be political theatre, however satisfying it will be to see the Clintons and their allies on the gallows. We have to aim higher if we really intend to drain the swamp.

These next few are taken from CHSmith’s end-of-the-week Musings report :

This is likely true, providing housing to the homeless reduces social costs, because it is just another aspect of preventative maintenance. Expanding the thinking, we would wrap everyone in enough social contacts, all kinds of clubs and interest groups, church groups, self-help groups, … that we would know about problems in our society locally and be able to prevent some of those causes or at least ameliorate their effects. A preventative maintenance society has to be very local :


More anti-Russian hysteria in propaganda from the WaPo, eagerly repeated by the NYTimes and many other propaganda outlets :



No kidding. While being an engineer in a research environment, I saw MD-Ph.D. and Ph.D. students working very long hours for little or nothing, and know 9/10 of them did not succeed in academic careers. In the work environments I have been in, a Ph.D. is necessary for some research jobs, and otherwise just serves as a measure that once you were intelligent and worked at one task long enough to finish a dissertation, but you still get interviewed same as any other people. I have hired several Ph.D.s as a manager, interviewed others, and worked with many others, so know that side of things. Many Ph.D. physicists end up as system programmers or firmware programmers or logic designers, people working at the hardware-software interface, requiring knowing several different technologies. On different projects, I hired a Ph.D.s and a couple of MSs in English Literature as writers (they can write), Ph.D.s in Experimental Psychology as programmers (they all built apparatus and did data analysis), … Researchers have practice in digging and demonstrating understanding by doing, crucial engineering skills :

View story at Medium.com

More background for understanding Pizzagate. When homosexuality was severely repressed by public opinions, men needing such were prey for blackmailers from all walks of life.  This recounts the story of a 10-year blackmail ring and how big the FBI and NYPD investigations were. Now it takes paedophilia for blackmail, and the blackmail continues, we now are more aware of it from the official police and intelligence agencies than from ordinary criminals :


It is so hard to choose, there are so many, so egregious and so frequently in the news. There are many more of every kind, so frequent they go to page 3 and below, never read :


Lovely photos of Pakistan :



And another few links away is this, more evidence that the Persian-influenced cultures are deep and nuanced in all things. The Muslim religion has just as varied a history of interpretations as has had Christianity. Our education systems are careful to hide all of this from everyone :


And this :


Georgian Queen Gürcü Hatun was a patron and a close friend of Rumi. She was the one who sponsored the construction of his tomb in Konya.[49] The 13th century Mevlâna Mausoleum, with its mosque, dance hall, dervish living quarters, school and tombs of some leaders of the Mevlevi Order, continues to this day to draw pilgrims from all parts of the Muslim and non-Muslim world. Jalal al-Din who is also known as Rumi, was a philosopher and mystic of Islam. His doctrine advocates unlimited tolerance, positive reasoning, goodness, charity and awareness through love. To him and to his disciples all religions are more or less truth. Looking with the same eye on Muslim, Jew and Christian alike, his peaceful and tolerant teaching has appealed to people of all sects and creeds.

Israel’s long propaganda campaign against Muslims has produced intolerance on all sides, and the world will pay for that until we all learn that humans are one species, rather inbred, and none superior in any permanent way.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s