from the what-an-entangled-web-we-weave? dept.
Physicists have developed a "physics shortcut" that allows ordinary laptops to solve complex quantum dynamics problems, a feat previously reserved for supercomputers and AI models (Live Science). The breakthrough, from the University at Buffalo, is an extension of a decades-old method called the truncated Wigner approximation (TWA).
TWA is a semiclassical approach that simplifies quantum math by retaining necessary quantum behavior while discarding less critical details. Historically, applying TWA required re-deriving complicated math for every new problem, making it inaccessible. The team transformed this into a user-friendly "conversion table" that translates a quantum problem into solvable equations, allowing physicists to get usable results on a consumer laptop within hours (University at Buffalo).
This new, practical approach significantly lowers the computational cost and makes exploring certain quantum phenomena much easier. It's hoped that this will save supercomputing resources for the truly intractable quantum systems, while allowing more common quantum dynamics to be studied efficiently on accessible consumer-grade computers (ScienceDaily).
(Score: 2, Touché) by Anonymous Coward on Tuesday December 09 2025, @01:42AM (1 child)
pi = 3
(Score: 0) by Anonymous Coward on Tuesday December 09 2025, @10:19PM
Teach the controversy!
(Score: 3, Interesting) by anubi on Tuesday December 09 2025, @02:17AM (23 children)
I remember Fuzzy Logic going around. It appeared to me they were trying to emulate an old-school analog computer with a digital one. I didn't get into it much as I had gotten pretty good at designing analog-digital hybrid PID circuits, having each do what it did best. I could make dandy little control systems with the analog part doing the P and D function while a digital counter, comparator, and sigma-delta type circuits handled longer term error correction, as analog circuits were prone to drift big-time. By using both, I could make both extremely agile as well as extremely precise PID type controllers using simple microcontrollers and inexpensive garden-variety jelly-bean analog parts.
I get the idea they are doing something similar using complementary technologies.
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 5, Interesting) by JoeMerchant on Tuesday December 09 2025, @03:51AM (17 children)
I made an autopilot for a (hobby sized) small fixed wing drone system. While the basic control systems in it were PID (with virtually zero D, and very small I), I did implement step-wise linear P values that are sort-of fuzzy-logic-like. Near target, P is low, and you hope the small I tweaks you around to where you want to be. Ordinary maneuvering takes you out into "normal" P range which works pretty well most of the time, but... when things are getting out of the normal operating range, I preferred for the control surfaces to get up to max deflection sooner than "normal" P would get them there, so there was a third range where P ramped steeply to achieve a rapid return to normal range - this happened to the small plane when it would hit turbulence that would knock it off kilter. In theory, you can tweak the D values to get that effect, but I had better luck (less unwanted control maneuvers) with both the real toy plane and the Cessna in the Flight Gear simulator using the stepwise linear P response.
Around about 1986 I had a professor who was really jazzed about fuzzy logic, I just didn't see the appeal of the formalizations - it all looked incredibly obvious to me, and also seemed to demand simulation testing to check for proper results - no real formal methods to predict what's going to happen, just tweak it then try it and see if it does anything stupid (aka unanticipated.)
🌻🌻🌻🌻✌️ [google.com]
(Score: 5, Interesting) by khallow on Tuesday December 09 2025, @04:37AM (8 children)
From the fuzzy logic point of view, one of the interesting aspects of such prediction markets is their skew or "correlation". If all claims were independent, then the probability of any Boolean "and" chain of claims or their negations would be the product of the probabilities of each piece in the chain. If A has probability 1/3 and B has probability 1/5, then "A and not B" has probability of 4/15.
But they aren't independent and much of the market was about figured out how these were correlated.
(Score: 2, Insightful) by anubi on Tuesday December 09 2025, @05:08AM
Multiplication by sigma-delta streams...which I saw as just a stream of ramp up or down for the duration of a clock interval. Continuous up or down meant input was out of bounds. Like you indicated, fractions, reciprocals, and multipliers were done with gates. I thought that what all that excitement over fuzzy logic was all about. Let it run long enough and it would asmtotically arrive at its solution with a minimal amount of hardware.
But it wasn't fast. When it got there, it would just buzz around the answer.
It was handy for some processes that had a long physical time constant.
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 3, Interesting) by JoeMerchant on Tuesday December 09 2025, @01:30PM (6 children)
> If A has probability 1/3 and B has probability 1/5, then "A and not B" has probability of 4/15.
I still have a problem seeing that as anything other than formalization of the obvious and putting the "fuzzy" label on it for no apparent reason. What else would the net probability be if it didn't have the "fuzzy" label?
My professor was all jazzed about applications in things like washing machines: how long do you soak? It could be a fixed value of 5 minutes, but with fuzzy logic you might make it 6 minutes, decreasing to 3 minutes depending on the water temperature, any water temp above 30C would decrease the soak time on a linear ramp (6 minutes at 30C and below, 3 minutes at 60C and above, linear slope between). Then you could complicate it further with other factors, but he never really produced any practical examples of more complex logic for washing machines. I think he proposed toaster ovens might adjust toast time based on both ambient temperature and humidity although that seemed a bit of a stretch and questionable whether the two inputs were doing anything that a single input of wet-bulb temperature wouldn't do by itself, or anything practical at all given the impracticality of sensing the state of the object to be toasted and the relative irrelevance of ambient temps and humidities vs the toasting environment.
🌻🌻🌻🌻✌️ [google.com]
(Score: 1) by khallow on Tuesday December 09 2025, @01:39PM (3 children)
What's obvious about probability? And linguistically, how can you talk about something, if you don't have labels for the concepts?
(Score: 0) by Anonymous Coward on Tuesday December 09 2025, @02:29PM
That it's "close enough"
You make them up. Advertising, babe
(Score: 2) by JoeMerchant on Tuesday December 09 2025, @04:50PM (1 child)
>What's obvious about probability?
It's obvious to me that if you need two things to happen, like a coin flipping turning up heads twice, that the cumulative probability of those things is the product: 0.5*0.5 = 0.25. And that this concept extends to any set of probabilites, like 3D-20 coming up 1-1-1 = 0.05 * 0.05 * 0.05 = 0.000125
How that has anything to do with the linguistic label "fuzzy" is mysterious to me.
🌻🌻🌻🌻✌️ [google.com]
(Score: 1) by khallow on Tuesday December 09 2025, @06:41PM
But if that coin was used earlier roll a bunch of heads, then it used up its luck and will roll more tails. Unless you get a properly trained Boston terrier to lick the coin first and put the good luck back in.
Gotta call it something. I think "fuzzy" came about because it was a representation of a mixed state rather than a pure state. Things described that way are naturally fuzzy.
(Score: 2) by acid andy on Wednesday December 10 2025, @09:14PM (1 child)
Isn't that just what they call heuristics?
"rancid randy has a dialogue with herself[...] Somebody help him!" -- Anonymous Coward.
(Score: 3, Interesting) by JoeMerchant on Thursday December 11 2025, @01:19AM
>Isn't that just what they call heuristics?
Yeah, I wasn't impressed and for my elective project in his class I did an working example of a Shannon 4 from 7 error correction encoder / decoder - buttons, lights, wires, it all worked, including leds showing you the faulty line if there was one, but he wasn't impressed, that got me a B for the class, B- on the project "old stuff from the 50's, bah." Meanwhile, anyone who turned in at least half a page of scribbled lines with the word fuzzy in there at least 5 times got at least an A for the class and project.
🌻🌻🌻🌻✌️ [google.com]
(Score: 3, Interesting) by VLM on Tuesday December 09 2025, @06:46PM (1 child)
I'm of a similar vintage with similar interests, and the way I recall it was the fuzzies had like a prayer to the gods of the free market, if we formally verbally pray hard enough for formal verification systems for fuzzy logic then they'll magically appear via the marketplace of ideas.
That did not happen; rather like wishing really hard for anti-gravity or warp drives. With predictable results on the fuzzy logic field, which kind of went away.
I like going to used book stores because you can pick up stuff like this all the time. Was not cool enough to be kept and scanned online until today in 2025, but is cool/retro to pick up for $4 and reminisce about the old days.
Another thing I remember about the "fuzzy logic bubble" of the 80s was people trying REALLY HARD to implement stochastic annealing the hard way without using the phrase stochastic annealing because thats tired and old fashioned whereas "fuzzy logic" is like total Star Trek cool and new. All stochastic annealing is, is to emulate an analog task numerically probably with some kind of optimizing goal, then intentionally ruin the analog loop by injecting a ton of noise, then smoothly gradually sink the added noise to zero and that'll usually find the global min/max instead of getting stuck, or at least it usually works better than competing algos.
Fuzzy logic was also incredibly inefficient, which probably didn't help acceptance. I remember a Bob Pease magazine column from the 80s where some dudes used like an entire DEC minicomputer or something to use fuzzy logic to balance an inverted pendulum using tons of 1980s CPU and floating point accelerators and ADCs and DACs and a pile of math to teach itself how to balance, so he made a pendulum balancer using about 6 op amps and some string and it worked better. Bob Pease was the guy who died driving home from Jim Williams funeral... crazy two pillars of analog electronics dead at the same event, kind of. IIRC Williams wrote like half of LTC's app notes, at least half of the good ones anyway. Pease's magazine columns were pretty interesting reading back in the 80s.
(Score: 3, Interesting) by JoeMerchant on Tuesday December 09 2025, @07:35PM
> if we formally verbally pray hard enough for formal verification systems for fuzzy logic then they'll magically appear via the marketplace of ideas.
That sounds about right...
>predictable results on the fuzzy logic field
Taking your statement completely out of context... I have been saying lately that the "power" of LLMs resides in their non-deterministic nature (fuzziness, if you will...) A deterministic computer program succeeds, or fails, and given the same inputs will have the same outputs every time. When you want what it's giving, then this is good. But what about when you don't know exactly what you want? The power of these "moderately warm temperature" LLMs is in large part because they are non-deterministic, they will give different answers for the same inputs, so the Beckett quote applies: try, fail, try again, fail again, fail better next time. I see this with LLMs doing a code refactor (partially cleaning up a little of the mess they made), 6/74 unit tests failing... Oh, I see the problem - blah blah - 2/74 unit tests failing... Oh, I see another problem - blah blah - 1/74 unit tests failing... and then the classic lie: 73/73 unit tests passing, one failing unit test from a pre-existing condition unrelated to my recent code changes... (yeah, except 74/74 were passing before you started...) Gotta watch your robots, they learn their reporting skills from people.
I read quite a few Pease columns back in the day, couldn't tell you anything specific about any of them just off the top of my head, but I'd bet he was fairly influential in my life anyway.
🌻🌻🌻🌻✌️ [google.com]
(Score: 1) by khallow on Saturday December 13 2025, @09:04AM (5 children)
These give the right truth table values when x and y are restricted to 0 and 1.
It gives you some weird stuff. x and not x is now x-x^2 rather than 0 (though the polynomial still has zeros at x=0 and x=1). This is relevant because in practice that usually should be zero (depending on the application).
(Score: 2) by JoeMerchant on Saturday December 13 2025, @02:59PM (4 children)
I think your fuzzy logic is a bit different than the one my prof was pushing in 1986.
In his world:
X = 0 while condition 1 (say X=1 while temperature 60
Then:
Y = 0 while condition 2: time 6
Now:
use X * Y >= 0.5 as a test to exit the spin state. Get sheepish about how it's all very complicated and needs more care than just a quick example on the chalkboard. As a student, be amazed that they're charging $1000 per credit hour to listen to this clown and get credit... suggest that X might be 0.5 while below their minimum conditions and range up to 1 by something like X=0.5 + 0.5 * (temp - 20) / 40... wrinkled brow on professor followed by "yes, that works for algebra, but this is Fuzzy Logic, this is different, that's not the PROPER FORM!" Learn a bit about human nature in exchange for my $3K tuition (thankfully, in my case, paid by the state for this class...) Watch professor briefly attempt to achieve THE FORM with X=(temp - 0)/40, ask what happens when temp 20, "it's all very complicated, class dismissed!" (10 minutes early.) THE FORM actually sort of works with temp/40 but still needs the case of x=0.5 while temp 20 and that yields a different shaped response curve than he was looking for... his PROPER FORM always ramped from 0 to 1 over a range.
🌻🌻🌻🌻✌️ [google.com]
(Score: 1) by khallow on Saturday December 13 2025, @03:38PM (3 children)
(Score: 2) by JoeMerchant on Saturday December 13 2025, @04:38PM (2 children)
Apologies, I keep forgetting that SN editor makes > and symbols disappear if you just type them in the editor without special escaping and codes, you get the general drift of the message though - nutty professor put Fuzzy Logic on his resume, studied one book for 15 minutes then tried to teach a class on it. He was a Chinese import - but not as priceless as the later Chinese import professor who never gave any male student less than a B grade and never gave any female student higher than a D grade, for four semesters before the administration finally told him "you can't do that here, anymore" without doing anything about the honors level female engineering students' whose GPAs got a huge dent from his Ds.
🌻🌻🌻🌻✌️ [google.com]
(Score: 1) by khallow on Saturday December 13 2025, @05:05PM (1 child)
(Score: 2) by JoeMerchant on Saturday December 13 2025, @05:33PM
One final rant on the editor, I typed the symbols > and < and previewed them, and they looked correct in the preview, but the editor subbed them back to the actual symbols, so looking at the preview and clicking submit still gave me the disappearing < ... a minor quirk, yet astounding how many years it persists in the code...
🌻🌻🌻🌻✌️ [google.com]
(Score: 2) by RS3 on Tuesday December 09 2025, @04:53AM (4 children)
All very interesting.
Did you ever look into "chopper stabilized" https://www.electrical4u.com/chopper-amplifier/ [electrical4u.com] circuits?
(Score: 1) by anubi on Tuesday December 09 2025, @05:18AM (3 children)
Oh yes. Even used some mechanical ones back in the day. Used optical ones too with spinning mirrors for directing light through reference and sample spectrographic analyses of liquids and gases. I can make very stable AC amplifiers. But DC amplifiers, uh, not so good. Couldn't tell what was drift, what was signal. I had to have that "carrier" reference...like a "lock-in" amplifier.
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 2, Interesting) by anubi on Tuesday December 09 2025, @05:55AM (2 children)
Yeh, the really neat thing about the chopper design is that it only uses one amplifier, so it can have even severe DC drift yet still have excellent DC stability, but what the amplifier has to have is be linear, low noise, and stable AC characteristics. The differential DC signal being measured is now seen as a peak-to-peak AC signal at a given frequency, You can now use AC amplification and bandpass techniques to achieve almost unimaginable sensitivity to this slightest imbalance between its two input signals ( say, a thermocouple, or voltages developed as metals corrode ).
A multiplexer input switch looking at two or more signals - then a really high gain narrowband AC amplifier, then a synchronous ( to the input switch ) demultiplexer switch to separate the time slots on the carrier frequency , a few sample/holds, and you have a very precise measurement of the difference in voltage of the inputs with almost unlimited gain, and quite insensitive to the internal DC drifts of even its own core amplifier,
Measurement scientists love these things.
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 3, Informative) by VLM on Tuesday December 09 2025, @07:23PM (1 child)
To agree with and extend your remarks the 1/f flicker noise at a couple dozen KHz is way lower than down near DC.
The "voltage noise density" of plain old opamps always ramps up an low input frequencies... unless you use a chopper and then magically the graph can be near flat.
Its not a large conceptual jump from a chopper amp to a switched capacitor audio filter. Those things are pretty nifty. Or a lock-in amp. I had an EE lab a long time ago (like decades ago) where they tried to teach us something about choppers and lock in amps and mainly we learned lock in amp appliances are really cool and really expensive. Your typical box from Stanford Research Labs that costs as much as a car but has UFO like performance stats until you learn the trick of how they work and even after that they're still cool. SRL sells (sold?) a lot of strange and expensive boxes like that.
For a variety of weird physics reasons sometimes "things" generate the same amount of noise power per octave, that'll give you the 1/f pink noise. And some of those "things" are in the innards of opamps so you're pretty much screwed if you want ultra low frequency ultra low noise performance from an opamp. However, what if you did the radio/RF people thing and had an IF amplifier chain at a totally sane and normal frequency and then upconverted/downconverted at the inputs and outputs thats a chopper amp more or less. Thats another way to look at choppers. If you don't like the performance of an opamp below 1 Hz, who cares upconvert the signal to 10 KHz do all your low noise amplifying there, then downconvert back to 1 Hz.
Around 1990 Analog Devices patented building a chopper with a spread spectrum clock to reduce unintended clock spurs in the output. AFAIK nothing came of that, at least commercially, and the patent expired. IIRC the overall noise floor went up a bit but the discrete spurs basically disappeared below being measurable, maybe an overall net win sometimes. One of those "interesting things to think about in my infinite spare time" topics.
(Score: 1) by anubi on Wednesday December 10 2025, @12:36PM
I have a Princeton Research 124A w/differential plugin I got at a company research sale, but never have had an opportunity to learn how to use it. It looks more like a sound board it's got so many dials to set up.
It's been a pet imaginary design of mine to build a colorimeter to look at those color changing reagents on soil test kits using various colors of LED synchronously driven, photo detector sensed, AC amplified, synchronous detection, using an Arduino to fit curves, for checking soil chemistries and logging same to a cellphone via configuring the Arduino as a Hotspot. I would be flashing various colors of LEDs looking for ratios of reaction.
Just walk around with it, take lots of samples, then end up with a GPS map colored on top of a downloaded Google map showing various indicating reagent results. LaMotte type indicator reagents.
It's something I think about, but it's one of those things that is too complex for me to do by myself right now. This thing could take me a couple of decades to build. And I am already swamped in projects I start and don't finish.
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]