TauZero

Needed a replacement 700C front wheel for my commuter bike after the old aluminum rim exploded like a looney tunes cannon. It's hard enough to find the right size when there are 3 competing tire/rim sizing systems currently in use, it doesn't help when the people selling the wheels have no idea what the numbers mean either! All of these examples are from separate storefronts at the big online store. Ended up buying the wheel identical to mine from my local bicycle shop at the same price as online and with no shipping fee or delivery time.

The 3 systems in use are the American customary inch fraction notation like 26x1+3⁄4 (which is NOT interchangeable with American customary inch decimal notation like 26x1.75), the French metric notation like 650x45C, and the ISO 5775 metric notation like 47-571. I found the wikipedia conversion tables and Sheldon Brown's tire size chart invaluable.

To better understand how neural networks function, researchers trained a toy 512-node neural network on a text dataset and then tried to identify features within the network that are semantically meaningful. The key observation is that while individual neurons are difficult to attribute specific functionality to, you can find groups of neurons that collectively do seem to fire in response to human-legible features and concepts. By some metric, the 4096-feature decomposition of the 512-node toy model explains 79% of the information within it. The researchers used an AI nicknamed Claude to automatically annotate all the features by guessing how a human would describe them, like for example feature #3647 "Abstract adjectives/verbs in credit/debt legal text", or the "sus" feature #3545. Browse through the visualization and see for yourself!

The researchers called the ability of neural networks to encode more information than they have neurons for as "superposition", and single neurons being responsible for multiple, sometimes seemingly unrelated, concepts as being "polysemantic".

Full paper: https://transformer-circuits.pub/2023/monosemantic-features/index.html
also discussed at: https://www.astralcodexten.com/p/god-help-us-lets-try-to-understand
and hackernews: https://news.ycombinator.com/item?id=38438261

OC just for you! ♥️

The recent post made me fear that a lot of you are taking this "monkey looks at double-slits" meme, which was only ever supposed to be a funny monkey meme, actually seriously. Honorable mention goes to @kromem@lemmy.world, whose 12 posts on the topic, insisting that the quantum eraser experiment (but not the delayed-choice quantum eraser!) proves that the double slit is somehow bizarre, forced me to make my own meme. This meme explains the (non-delayed choice) quantum eraser paper from arXiv:quant-ph/0106078 and the figures are numbered to reference the paper.

First of all, looking at the photons, you the conscious intelligent monkey, MAKES NO DIFFERENCE. You can't actually "see" the photons going through a slit the way you could see say a bowling ball. The only way to detect a photon is to absorb or reflect it, and if the photon is getting absorbed by your eye that means it's not going through the slit or hitting the screen. The interference pattern stays visible on the screen WHETHER OR NOT YOU LOOK AT IT.

They've lied to you when they said the pattern changes when you "look" at which slit it the photon goes through. What the physicists actually do to measure the "which path" information is they put these circular polarizer filters in front of the slits, one clockwise one counterclockwise. Then the pattern disappears and you get this one single blob of density (Not even double! Figure 3). This is because light polarized in opposite directions cannot interfere with itself - wikipedia calls this the "Fresnel–Arago laws". In principle you could have put a polarization detector in place of the screen and record which way the light hitting it is polarized, which would tell you which slit the photon must have went through. The physicists DON'T EVEN BOTHER DOING IT. The fact alone that the light is polarized when it hits the screen is sufficient to destroy the interference pattern.

Well, NO SHIT. You put these giant 3D glasses in front of the slits and you still expect to see interference? This is very much a "mechanical interaction", not some "non-obtrusive conscious observation". Everything that destroys coherence will ruin your quantum experiment! Mystery solved!

So what about the quantum eraser, @kromem will ask? Popular science has created this myth that you can look at the screen and you can make the interference pattern literally shimmer in and out of existence by just flipping a switch, connected to second detector positioned elsewhere, turning it off and on. An action at a distant place (the detector POL1 observing "twinned" entangled photons created by this fancy nonlinear barium crystal before the slits, Figure 1) changes whether light over here behaves as a particle or a wave, right in front of your eyes. Spooky action at a distance, right?

THIS FUCKING DOESN'T HAPPEN. The monkey will see the single blob from Figure 3 and only single blob, no matter whether it turns the second detector on or off! The interference pattern will NEVER shimmer back into existence. The light never switches between behaving like a wave and behaving like a particle. It always behaves the same way, all the time, everywhere in the universe - like fucking light!

So what do the physicists actually fucking mean when they say the interference pattern is "restored"? If you observe the photons hitting the screen one at a time and you correlate them with simultaneous detections at detector POL1, you can mark those events as either "yes coincidence" category A or "no coincidence" category B. If you look at just all the category A events (Figure 4) you will see an interference pattern, and just category B you will see another (Figure 5). You cannot see these patterns by eye on the screen! You have to use a computer to record the events individually and separate them, you will only ever see a single blob by eye. The two interference patterns are subsets of that blob. They were always part of it, their hills and valleys mesh together into a single continuum. NO ONE EVER FUCKING EXPLAINED THIS.

The detector POL1 has a linear polarizer filter in front of it, so straight out the gate it will not see 50% of the twinned photons at all, because they will get stuck in the filter. Your category A can never match more than 50% of events. It gets worse, since the non-linear crystal in reality has very low efficiency and most photons going through are not twinned, so you cannot measure category B directly. In the experiment they do it by rotating the filter 90°, which changes the correlation to category B. In the meme I show them as if the crystal was 100% efficient.

The delayed-choice quantum eraser works similarly - you only ever see a single blob and can never see the interference pattern shimmer in and out of existence. You need the correlation data from the second detector to split the blob into two intermeshed interference patterns using a computer. The Sabine video was the first one I've ever seen that explains this correctly. Every other popular science video up to that point has lied to me!

Whatever you do, DO NOT watch the DR. QUANTUM video with an open mind! (Not even going to link to it, @manual3204 linked it in the other thread.) It's from a documentary produced by a literal UFO cult to promote their quantum woo woo, only masquerading as a quirky science video. It came out in the early days of youtube, when its production and animation quality were unusually high for its time, so it immediately became youtube's go-to video for double slit experiment. Copies of it remain highly ranked there even to present day. It's total baloney!

All the McD*nalds in my area have been upgraded with order kiosks. Regardless of all the controversy around self-checkout, and minimum wage, and automation taking our jobs, I personally love them. I can take my sweet time composing my order, I can see the full selection (such as it is), I can see pictures and prices clearly without having to strain my eyes to read 12pt font on the tableau, and I don't have to shout at the cashier to be understood or struggle to hear back. I really believe this is the right way forward.

My only complaint so far has been that the order kiosks only accept card. There is actually a way to pay by cash that the machine never lets you know about - you have to press "cancel" on the keypad when it asks to insert card, and then the screen gives you an order number to give to the human cashier (each store still has one register open) so you can pay in cash. So I still have to wait on line, but at least my order selection is locked in, I can have exact change ready, and there isn't usually a line anyway anymore.

I know all yall Europeans are proud about your nearly total transition to cashless economy or whatever, and you like to boast how not a single euro banknote has graced the inside of your wallet in months. However I personally like cash, and I genuinely believe that a cash payment system is a necessary element of a liberal democracy and secure society. So at least understand my pleasant surprise when I saw these reverse-ATM cashboxes at this restaurant. They work and were being actively used too! (It spat out my dollar coins though, those bastards!) I hope they find their way into more places.

It is said that ACs are counterproductive in fight against global warming, in that while they may make the local environment temporarily livable, the greenhouse gases produced while making the electricity needed to operate them heat up the rest of the Earth by much more than the relief from the AC itself. By how much exactly is that? Note that here I am interested in the global impact of greenhouse gases specifically, not in the local heat island effect (given how ACs do not destroy heat but only move it from inside to outside, and add extra heat from running the compressor itself). Let's also assume all electricity comes from fossil fuels (ACs might become a viable solution if 100% of AC electricity came from renewable solar, which is actually a reasonable goal to strive for given how both AC and solar are most active during the day, but at the moment most of electricity delivered to me specifically, for example, comes from natural gas.)

Here's my estimate. Let me know if it is reasonable! Methane has energy density of 891 kJ/mol, burnt into CO2 at 1 mol : 1 mol. Gas turbines have efficiency up to 60%. The radiative forcing of CO2 can be calculated as: ln(new ppm/old ppm)/ln(2)*3.7 W/m**2. For example the 131 ppm increase in CO2 since 1750 up to 411 ppm has a radiative forcing of 2.05 W/m**2 (is that across the entire Earth's surface? or only its crosssection?), and CO2 has persistence in atmosphere for at least 1000 years. The atmosphere composition is 78% nitrogen 21% oxygen 0.9% argon so its molar mass is:

.78 * 28 g/mol + .21*32 g/mol + .009*18 g/mol = 28.7 g/mol 

And total atmospheric mass:

4*3.14*(6.37e6 m)**2 * ~10000 kg/m**2 * 1000 g/kg / (28.7 g/mol) = 1.78e20 mol

Suppose 8 billion people each run 1kW AC for 1 year, with electricity from natural gas. (That's similar to our total current global energy consumption of 20TW, though of course we use power for things other than just AC or electricity, but also most energy comes from coal and gasoline not just gas, and 80% comes from fossil fuels not renewables.)

8e9 people * 1000 W/person * 60*60*24*365 s / (891e3 J/mol * 0.6) = 472e12 mol

That's 472 teramols of CO2 (20.8 gigatons) added to the atmosphere each year, or 472e12 / 1.78e20 * 1e6 = 2.65 ppm (parts per million). It is believable that having done so for a hundred years we have raised CO2 concentration from pre-industrial levels up to 411 ppm. The radiative forcing is:

ln((411 ppm + 2.65 ppm)/(411 ppm)) / ln(2) * 3.7 W/m**2 = 0.0343 W/m**2

Or for the whole earth:

4*3.14*(6.37e6 m)**2 * 0.0343 W/m**2 = 17.5 TW

What is my individual contribution for 1 hour?

17.5e12 W / 8e9 / (24*365) = 0.25 W

That is, if I run my 1kW air conditioner for 1 hour, the entire Earth will be solar heated by an extra 0.25 W for the next 1000 years. That doesn't sound like much, but it adds up over time: I spent one kilowatt-hour in one hour on cooling, but the rest of the Earth will be heated by an extra 0.25 W * 24*365 hours = 2.2 kilowatt-hours in the next year, and again every year thereafter. Multiply that by 8 billion people or a hundred years and it adds up a lot, even considering the heat is distributed across entire planet surface not just areas where people live.

So my answer is 1 kWh of cooling = 2.2 kWh of heating per year for the next 1000 years. By same calculation in terms of mass, 1 kg of CO2 = 7.4 kWh of heating for every year thereafter. Is this accurate?

Everyone is armed all the time and that's normal, but to draw a weapon is an overt hostile act. A standoff therefore is a game of chicken because both want to kill each other and you want to draw first to have the highest chance of surviving, but even a bandit will hesitate to add a felony murder charge to their rap sheet. The whole town serves as witness when there is a pair of eyes behind every shuttered window. The hero always draws second, both demonstrating his superior skill and speed by defeating the opponent even at a disadvantage, and getting away with murder scot-free.

Image description - Cuphead Rage Flower meme
Me seeing cheap and expensive nuts mixed together for sale on the shelf
Me buying expensive nuts and mixing in cheap nuts myself

[Edited for CAPSLOCK]