robryk @robryk@qoto.org

A colleague mentioned today that the April Fools tradition of pranking unsuspecting people into believing something false can be very unwelcome on the receiving side. That made me think of better ways to observe April Fools and I think I've found a slightly Discordian one that I wish I'd thought of years ago.

Let's share puzzles/riddles that often leave the listener very confused and help them realize that something they might have believed about the world is inaccurate. I think it's much better, because it's educational, there's no temptation not to ask the recipient whether they wish to take part beforehand, and I don't expect recipients to feel like they're being made fun of.

Let me start with a physics puzzle I'm fond of:

Consider a car that travels northward with speed v. Assume there are no losses (no rolling friction, vacuum, etc.) so the car travels at constant speed with engine off. At a point in time, the car engages its engine and speeds up to 2*v northward. How much work did the car engine do?

Well, we can compute the increase of car's kinetic energy: m/2*((2v)^2-v^2)=m/2*3v^2

Alas, let us consider a different (inertial) reference frame: one that moves northward with speed v (note that it's not tied to the car, even though it starts stationary in it). In that reference frame the car sped up from 0 to v, so the increase in car's kinetic energy is m/2*v^2.

What gives? How much energy did the engine actually have to use to speed the car up?

h/t to Ryszard Zapała, my HS physics teacher

#freecad I'm trying to design some pretty simple shapes using FreeCAD (using Part Design and Sketcher) and I keep encountering weird issues, including things that look like caching problems (I need to twiddle a value back and forth to see effects of other changes reflected) and outright crashes (https://tracker.freecadweb.org/view.php?id=4523 and another that boils down to "this file crashes FreeCAD if I try to edit ~anything in it").

I am most likely using it in a weird way. However, I didn't expect that amount of issues (esp. crashes) even if I was using it completely incorrectly. Is it expected that FreeCAD will be crashy-when-used-weirdly, or maybe is it the fault of my distro (NixOS), or something else?

If only I didn't want chamfers, I would just use solvespace. Unfortunately, chamfers on non-side edges of extrusions are very annoying there.

@freemo

> Except those numbers are based on completely fictious numbers with no basis in reality…

I admit I haven't tried to look them up now, these are just ballparks of lower bounds that I kind-of believe within the order of magnitude.

However, I realized that we should have a better way of estimating the risk of "dying from covid over a period of time": we know (for many areas) how many people were diagnosed with covid and died as its result (for some definition thereof). For example, ~1/1300 of Switzerland's population died of Covid since the beginning of the pandemic according to the official stats (https://www.covid19.admin.ch/en/overview?ovTime=total gives the number of deaths as 6003, population is ~8M).

> Imagine a worst case ADE scenario, for example… in a year after the immune system has settled and everyone is in a state where theya re on long-term antibodies rather than short, now everyone getting COVID all of a sudden starts dying at 10x the rate as before due to ADE.. what was meant as a vaccine has now turned the disease into a far more lethal killer (basically how ADE plays out)… while we dont have the data to know this will be so, and we are completely blind as to its possibilities, the fact is, that is one worst case that could be absolutely horrific if it plays out.

Aren't we similarly blind to the possibility of a Covid infection causing similar effects (á la Dengue fever)?

It seems to me that you're suggesting that in face of two alternatives with unknown outcomes we should default to doing nothing. Do I understand correctly, or are you using some more specific principle? If so, why should we consider the "do nothing" alternative specially preferred?

I believe that there's a good reason to treat "do nothing" as the preferred option when the alternatives are presented by an adversary (e.g. when someone gives us a "lucrative" offer under time pressure, a reasonable default action is to refuse; the "do nothing" alternative is the only one the adversary cannot fail to include). This obviously doesn't apply here and I don't think that outside of adversarial context this is a good principle.

@freemo @Science

> No vaccine is 100% safe, but if a billion people are vaccinated, a one in 10,000 serious adverse event will affect 100,000 of those people."

Do you know of anyone trying to compare such risks against risks of delaying?

If we assume a 1% chance of contracting Covid-19 in half a year, 0.1% infection mortality rate, and 50% effectiveness of the vaccine, the vaccination removes a 5*10^{-6} risk of death over a half a year.

This is 20 times smaller than the hypothetical 1/10,000 serious adverse event chance. At a first glance, this would indicate that for anyone who values life quality after a serious adverse event at >=95% of life quality now should get vaccinated without waiting for more data/interpretation thereof, assuming no better vaccine becomes available within half a year.

My estimates above are crude and make assumptions based on what I remember reading, so I'm pretty sure they are off, possibly by an order of magnitude. I also might be missing something qualitatively.

Do you know of any research trying to express vaccination costs/benefits in similar terms? Or maybe of a reason why doing it in these terms is not useful?

@Demosthenes It's a 4mm banana socket. Thus, there is ~no lock-in and ~universal compatibility between ones produced by different companies.

Thanks, I haven't considered the "get hobbyists to like our thing" angle.

@Demosthenes Are you saying that this strategy precludes the existence of such sanity checks, or that this might be an example of such a strategy?

I would doubt the latter in this case: the product costs ~1$, and you're really unlikely to be buying <10 unless you are a hobbyist.

How do shops end up with such anti-discounts for volume orders? I don't expect this is intended, so why isn't a sanity check against such cases a common thing?

@freemo @torresjrjr @Diptchip @Science

> Presuming that is, in fact, what you meant the failure here is that you can only determine how “true” a model is if you are omnipotent with regards to the problem, you already know the outcome. So in any practical sense it wouldn’t be a valuable way to interpret real world models.

You can approximate KL divergence by sampling from the "world" distribution (easiest way to see how: it's essentially expected value of log(1/p) where p is the probability the model assigned to the outcome we've sampled). That makes KL divergence estimable (with the small exception of models that assign probability of 0 to any outcome) when comparing models against the real world (insofar any estimates can be made against the real world).

> So to answer the question of if the coinflip model is a good one… Its an amazing model, if you are trying to model the chance one of two randomly selected people might win, without any other knowledge.
> It is a very poor model at predicting other things however, (...)

I agree completely. I would phrase it as it being the best model over the system where the outcome (white/black wins) is the only random variable being modeled.

> So while it is, clearly the better model, this is not due to it being a closer representation of the underlying system, it is merely better because it predicts the outcome more accurately.

I don't understand the distinction, or, perhaps I should say that I don't understand what "closer representation" means. Is it something that can be evaluated (even by an omnipotent evaluator)?

@torresjrjr @freemo @Diptchip @Science

Hmm.. I don't get the parable. Models' accuracies are evaluated not by comparing their average prediction with average result, but by something like comparing the "leftover surprise" the model leaves us with (KL divergence of the probability distribution of the world with respect to probability distribution that the model predicts). On that count, the model that models the intricacies of chess is obviously more accurate, and the question of whether the increased complexity is "worth it" is not answered obviously in the negative.

Consider the following example: let's say that we have a clock that has a dot that blinks, so that it's on during every even second and off during every odd second. The article, if I extrapolate correctly, would say that the model that says "at every point in time the probability that the light is on is 1/2" is just as _accurate_ (ignoring the question of its complexity) as the model that says "at every point in time the light is on iff it was off a second ago". I don't think that any model evaluation method that would claim that is useful. Thus, I see that article as, in large part, a strawman against a model evaluation method that isn't really used nor is intuitive. Is there something I'm missing?

@rysiek @timorl It seems that all values I can get the graph to show are rounded to integers. That makes sense for all of them, except for values per 100k/1M people. The quantization of those causes logscale per-population graphs to look very staircasey, see e.g. https://rys.io/covid/#delta,per100k,date;poland,switzerland

This is really hilarious: https://youtu.be/cuWVWUaNdqc?t=2374

Court arguments can be hilariously entertaining (because judges are very good at cutting through bullshit) and frustrating (because it's supposed to be pure eristics from both counsels and because many problems are caused by lack of precision in the law) at the same time. I just got entertained by a discussion about differences between kinds of delivery and about excluded middle: https://youtu.be/cuWVWUaNdqc?t=1486

What are the advantages of through-hole PCBs over wire wrapping?

I can see many advantages of wire wrap (though I've never used it, so I might be mistaken):
- due to wire shape, unwanted capacitances are kept much smaller in wire wrap (compared to e.g. a 2-layer PCB),
- it should be about as labor-intensive to wire wrap through-hole components as it is to solder them (TTBOMK you can't reflow through-hole components, so you need to spend some time per pin),
- no need for explicitly making multi-layer PCBs: the problems of routing, possibly stacking layers (for >2 layer ones), coating vias, etc. just go away.

The only advantage of through-hole PCBs I can see is that they're more durable mechanically and thinner.

Am I missing something obvious?

@freemo Ah, does your prediction of equal heating extend to the situation when there is ionized gas in the gap?

@freemo I'm thinking how to set up an experiment where I can tell if both electrodes would heat equally, because I still doubt that :) (my best approach so far is to buy and modify a fluorescent lamp by inserting a bridge after the inductor).

@freemo Voltage is the energy per unit of charge needed to move a charge between two points, no? If there's a potential difference of 50eV to overcome for electrons, I need to apply 50V across it to cause a nontrivial current to flow

I'm not sure why that analogy should fit the situation better than an analogy where the ball is constrained to the surface (e.g. it's on a wire that follows the shape of the potential).

@freemo This is surprising for me. I would expect the potential energy of an electron inside a conductor to be lower than that of a electron in vacuum, and thus would expect that this difference has to be overcome by the applied voltage (and then would expect the anode to be heated by the incoming electrons, when their potential energy drops). (I'm ignoring tunneling, which is why I added the macroscopic distance caveat.) Is there an obvious reason why this is not so, or am I thinking in completely incorrect terms?

@freemo The gap, assuming it's of macroscopic size, should have a V/I characteristic similar to diode's: as long as the voltage is lower than the depth of the conduction band in the negative electrode, no current will flow. If the voltage is higher, I'm not sure what would actually limit the current.

Do you expect the "threshold voltage" to be lower if the gap is filled with unionized gas? The threshold voltage for unionized gas is easily reached by starters of fluorescent tubes.

@freemo If they actually have nonzero resistance and are mirror images of each other, should they heat up equally?

If they're superconductors, what will be the voltage drop between them? If it's not zero, we're sinking that energy (current*voltage drop per unit of time) somewhere; where?