On miscellaneous lesser ideas

 0. Posts on this blog are ranked in decreasing order of likeability to myself. This post is a catch-all repository of miscellaneous lesser ideas that would otherwise clog the bottom of my preferences, with additions coming up all the time and each new leading number indicating a separate entry.

 

1.1 (20220630) What is the optimal amount of policing? Here’s a simple, reasonably realistic and entirely obvious model that is relevant to the question: the more freedom of action and freedom from scrutiny the police have, the more they are able to keep visible crime under control. The more controls are applied, the less are they able to keep criminals in check. On the other hand, higher freedom of action causes increasing abuses of power and allows police forces to act like gangs writ large which, unlike honest gangs, can hide behind the cover of the state.

1.2 In this model, the optimal amount of freedom and resources allowed to the policing effort is such as to allow the average citizen to be bothered by policemen about as much as he is afraid of crime, on the margin (those who think that it cannot be possible to be bothered by police more than or about as much as one’s afraid of crime have not lived behind the iron curtain).

1.3 Every now and then some technology may come online that allows us to lengthen the leash police are on without increasing the degree of bother to the community. As an example, when body cams were adopted, these curtailed the some of the worst excesses routinely alleged earlier, thus allowing the police to be given extra freedom of action and bring the equal bother scenario to a lower overall level. Street and street-facing cameras, smartphones, tasers, changing the racial mix of the population, etc.

1.4 So, you can’t make crime go away lest you install the dictatorship of the police, and you can’t defund the police lest gangs roam free. You balance them against one another, and hope that tech comes along to make the trade-off more bearable.

 

 

2.1 (20221119) Breeding with your first cousin (or worse) is a well-known recipe for the proliferation of a multitude of genetic disorders, a genetic dead-end that will eventually be the death of the species if carried out with obtuse insistence. But is there a limit on how dissimilar partners can be, as long as they are still of the same species and are able to breed fertile offsprings?

2.2 I conjecture that the more genetically dissimilar partners are, the less the chances for the offspring to acquire some average of the parent’s traits, instead of randomly going with one parent’s traits over another. Think intelligence or height. In other words, if two very dissimilar people have an offspring, he will end up getting much of his traits from one parent or another, instead of a mix of both. A clone almost.

2.3 Hence, my conjecture is that there is some optimal genetic distance between partners, such as to minimise the risk of recycling of genetic disorders whilst allowing true averaging of traits into a coherent and functional whole (this being the whole point of sexual reproduction and ultimately evolution).

2.4 This would be some nth degree of cousinship which, going by gossip and centuries old practices in Europe broadly and my homeland specifically, sounds like its about 6th or 7th. So, here’s a business idea: create a dating app that creates a pool of eligible 6th - 7th cousins.

3.1 (20221207) Drawing inspiration from cat bonds in the insurance world, here's an idea: war bonds. Govts would issue these standard bonds, which in case of formal declaration of war are cancelled, with the principal and any interest still outstanding remaining with the issuing government.  Ideally the principal is taken in a liquid foreign currency but interest is paid in your local currency. A neat way to ensure you have cash when in real need, at the price of paying higher than usual interest when you don't. Obviously, the contract would be such as to allow the triggering event only after a set number of years (say, five) have passed to minimise moral hazard. Alas, the design of these is such that they have to rely on balloon payments instead of being fully ammortized. 

4.1 Aggregated into a main post.

5.1 (20221222) Would there be any niche application for commercial aviation based on a glider or range-extended glider plane with a dozen or two passengers? A common sailplane with an enlarged cabin, launched by cables as a carrier plane is, and either coasting on air alone or having a very small jet just to bring its glide ratio above even the extreme figures common now. Probably best for routes under 100km where straight-line travel shaves enough of the otherwise convoluted route to make it worthwhile. I guess it would be cheapish to run?

6.1 (20221222) I imagine you can shave some of the prison-like nature of high-rise apartments and the cost it entails in terms of fertility by having only one apartment in the building’s cross section: one apartment only, with visibility across three sides, with the fourth being the common area needed for an elevator and stairs. No neighbors around, only above or below. I realize such a thin building couldn’t be as tall as otherwise, but just how tall can we make this? Enough to make it commercially viable?

7.1 (20230106) The different approach taken by Gohlke and Hanson to the issue of status ranking makes one think of a football or other sport competition run in tiered fashion instead of the round robin fashion. Competing teams are assigned to the same baseline tier on day one, and paired randomly within the tier. Victories promote you to the tier above, draws or losses demote you to the tier below. When a team is found at the topmost tier with no other competitors, they are declared the winner, which doesn't stop the competition. A perpetual competition would lack the drama of the deciding final but would allow an inordinate number of teams to compete in the same championship, say all European teams in one continent-wide tournament than never ends. I really want to know where Vllaznia would be ranked, you know?

8.1 (20230108) It is common to hear the lament of the low TFR, or how entire countries are headed for a very dark future on account of falling births. If you really pay attention to what is being said though, you will notice that the real issues being lamented are two, geriatrics (too many old folks) and disgenics (the negative correlation between economic competitiveness and children). 

8.2 I'll add a third, overpopulation: once you account for the first two, there just aren't enough jobs than even a hundred million coders could do, even if we had a hundred million coders. 

8.3 It is important to disentangle these three effects since the solutions to each tend to be inversely correlated with those of the other two, to a degree that lamenting "too few youngsters" cannot capture. 

9.1 (20230326) An interesting visualisation I can think of but lack the skills to code for: a heartland heat map for a given country. Take a country and some subdivisions you know the GDP and border lengths of. Do either of,

9.2 Find the smaller group of contiguous subdivisions that produce 50% of the whole's GDP whilst having the shortest external border. Repeat the operation on this area, and then again, until you reach the limits of your subdivision resolution. Or, 

9.3 Iteratively remove that borderline subdivision such that the remaining part maximises the GDP / external borders ratio. Keep going until, again, you reach the limits of your subdivision resolution.

9.4 In either case, color-code successively more core maps (which I can co) and always reinforce the borders of these in tree trunk-like fashion (which I cannot).

10.1 (20230328) An algorithm for allocating a budget across several entities / items across years, such that the entities have some revenue predictability but your allocator’s freedom is also retained: at each year, you can allocate to each entity / item no less than the least of a) what you allocated last year, and b) the percentage of last year’s total budged allocated last year, applied to this year’s total budget. 

11.1 Aggregated into a main post.

12.1 (20230328) I rely on Russian Roulette for so much of these posts, that I’ve given some thoughts to the design of a modern weapon to be used for such purposes. Actual revolvers are meh, so instead have a modern automatic pistol, with a mini-drum across which the hammer has to pass to strike the primer. Each space in the drum can be set to allow transfer of mechanical energy (spring-loaded pin) or not (empty). The drum can be fixed in place for the gun to function as a normal, slightly heavier pistol, or can be rotated to allow the gun to function as a Russian Roulette tool.

13.1 (20230526) Not terribly useful, but I've noticed that if you numerically solve problem by minimizing the sum of squares you get such a function that tracks the average of what you're trying to model. If you solve by minimizing the sum of absolute errors instead you get the median. And in general, if you solve by minimizing a * (error if positive) + (1 - a) * (absolute error if error is negative) you get the a-th percentile of the process you're trying to model. 

14.1 (20230716) It is said that the sort of IQ test we use fails at the higher end of intelligence. Here’s an idea on how to bypass this issue: run the currently-available test on a relatively large sample on guys and calculate the first two moments of the results. Then, assign everyone the IQ result that would fit with those moments (assume some distribution) and their rank among the sample. Bootstrap (resample many times with replacement) a subsample from this sample and repeat, and you will have a distribution of IQ estimates for each participant, some of whom will be estimated at a higher level than we credit our tests with being able to measure. 

15.1 (20230716) A thought experiment on a new type of auction: invite bids and allocate the object to a random participant at a probability directly proportional to their share of the total bid across all participants. You’d expect the sum total of bids to be meagre due to the insane degree if risk (failed bids do not get their money back), but it's still the sum total of bids instead of the winning only, and may exceed the latter. Also, what appears to be a humongous betting industry would give one hope that this’d be, in fact, a way to maximize the price of the object (or at least invite a sort of wisdom-of-the-crowd element to the auction). Further, it allows the sale to one winner of arbitrarily large and expensive objects. 

15.2 Now, to make this different from a common prize lottery (though I'd be interested to see what total revenue those pull in relation to the estimated value of the object), allow bidders to change or rescind bids before the final calculation, and give them a real-time probability of success. Worth testing at a small scale.

16.1 (20230716) Keeping with today’s auction theme, there are such times where a large enough number of people inherit an indivisible object such as to make coordination problematic. Would inheritance just functioning as a gateway to an auction (only open to the inheritors, with the proceeds equally shared among the "losers"), with one sole winner as usual otherwise work in such and other instances?

17.1 (20230803) Suppose you graph the degree to which your progeny allows your genes to live on a 2D graph: on the y-axis the familiar rule by which each child is 50% of you, and each grandchild 25% of you, and so on, summed over your descendants n generations from you. Maybe you graph each discrete nth generation.

17.2 On the x-axis the chance of each individual gene having survived into the nth generation. So, if I only have one child, any individual gene has a 50% chance of surviving in that generation. If I have two kids, its now 1- 50% * 50% = 75% and so no. Approaches one, but never reaches it, not even for Genghis Khan.

17.3 This graph is not particularly informative one generation from you, since for every number of kids you have one and only one combination of y and x coordinates. But once you get into gen 2, things get interesting. A guy who has five grandkids through three kids (2, 2, 1) has coordinates of (if I haven’t stuffed up my math) 70.703125% and 1.25. But a guy having the same 5 grandkids through two kids only (3,2) has coordinates of 64.84375% and 1.25. The divergence is due to the fact that the total chance of your genes carried by one particular kid cannot be more than 50%, and indeed cannot even be 50% except approximating infinity.

17.4 The first guy has diversified his genetic portfolio, whilst the second guy has (by virtue of not having had the third kid) already lost 87.5% - 75% = 12.5% of his genetic patrimony forever before any grandkids ever start being born. And their kids may loose 12.5% * 50% of your patrimony by having two instead of three kids, and so on.  

17.5 Now, both coordinates get complicated really quickly if you account for your siblings’ and cousins’ siblings, but the principle remain the same. 

18.1 (20230927) I've written here before that excess senility is the truly huge issue of our age, and will require some sort of solution going forward. Here's a quick and dirty thought on how to both limit life expectancy and simultaneously improve the quality-total of year of old age: allow seniors to use and abuse any and all drugs or other recreational substance with no limitation whatsoever. I know this will leak like a sieve into the general population and will make control efforts much harder, but I think its worth it. One has toiled and looked after one's self one's whole life, and it's probably fair that one should be allowed to throw caution to the wind after a certain point.

19.1 (20231008) Another quick and dirty immigration algorithm. Would-be immigrants are - after being subject so some sort of filter - accepted as permanent residents, under the condition that two generations (starting with all who enter, regardless of age) are to only have one child at most. The third generation thus born are born citizens, but the first two remain permanent residents, and no amount of marrying citizens can change that. 

20.1 (20240330) Children's soccer modified such as to disallow head use (basically the use of anything over armpit height) to minimise concussion risk. Probably will have a deleterious effect on entertainment value, hence why to be applied to children's games.

21.1 (20240402) If you ever need to impose orthogonality (no collinearity) between a regression's predictors and don't mind loosing any explanatory power due to the interplay of different predictors, just run a series of simple linear regressions, at each step running the predictor most correlated to the error of the previous regression. There, no VIF, no PCA fakery.

22.1 (20240402) And speaking of PCA, here's a (rather obvious) alternative to it: arrange the data on a plane such that the geometric distance between point matches as much as possible the higher-dimension distance between the original data, by having the largest absolute deviation between projection and real be as small as possible.

22.2 Probably very computationally intense, though one may speed this up by solving for three points only (for which a full solution exists) and then solve after adding one more point in succession. 

23.1 (20240910) A different way of computing the IRV winner: run an iterated Warren or Meek count on all candidates with a quota equal to votes/candidates, such as to give every running candidate the same votes exactly. Eliminate the one with a keep value of one, iterate on the remainder, and so on until you have one winner. Produces very different results from IRV.