I come to entropy from Machine Learning, Information Theory and probability. For me it's fairly straightforwad. Of interest and useful - but nothing mysterious there.
The p.d.f. aka a fancy histogram is my current best knowledge of how many times an outcome is expected to happen. When I do one experiment - I can't tell the outcome. I can only count the number of different outcomes, without being certain when will any of them appear exactly.
A flat p.d.f. means my knowledge is poor: every outcome is about equaly possible. I'm very ignorant (high entropy). A spiky p.d.f. means I have good knowledge: some outcome is much more likely (low entropy). In extremis the p.d.f. is a Dirac impulse - that's deterministic knowledge.
The only mildly interesting thing is when a new observation reduces my knowledge. Say right now I'm fairly certain I have not got cancer. My chances are 90:10 for my age. Tomorrow I take a test, the test comes back positive. Of people of my age that test positive, aboout half have cancer for real. After the test my chances are 50:50. Now I am perfectly ignorant whether I have cancer or not. Whereas before I took the test and got a positive result, I was very certain I have not got cancer. The new information (positive test result), transformed my probability of cancer from spiky marginal P_Y(y)={0.9,0.1} to perfectly ignorant conditional P_Y(y|+ve test)={0.5,0.5}.
ZX-Spectrum with 16K RAM cost £100. With 48K RAM it cost £130. The keyboard was bad. Basic commands were assigned to combinations of keys, not typed letter by letter. The cpu was Z80A at almost 4MHz!
Most popular, as it was very compact. Could be smuggled across customs hidden under the coat.
Usually from West Germany. :-) My then country (YU) had banned importing a computer more expensive than 50 DEM (25 Eur). Knew of many kids that had ZX-Spectrum.
Commodore-64 had 64K RAM (not all addresable at once), much better keyboard. It cost £200. Basic programs could be typed normally. Learned Basic and 6502 assembler on it. The cpu was 6510 at almost 1MHz, had few memory mapped io ports too.
My parents relented after 2 years of pestering and bought me one from an authorized dealer. They were not rebelious enough to risk smuggling of ZX-Spectrum clandestine op. :-) The excise and duty paid were astronomical. It cost them £470 or sth like that. Eternally grateful to my parents for that. It changed my life. Knew couple of kids with C64 in addition to 2 of my closest friends.
BBC Micro B was £400. It was exclusive, like royalty, I knew of only one kid that had it.
With Commodore-64 you also had to buy their tape recorder (same thing as with Atari 800XL), while Spectrum worked with any existing recorder, which every house already had at the time.
The advantage of the C64 tape recorder was it reliability and the motor being pilot by the computer. With personal tape recorders there was a lot of problems due to heads miss aligned.
As I remember from that time, all C64 users also had tiny screwdriver for adjusting head azimuth. Could be related to pirated tapes in ex Yugoslavia, copied on who knows what kind of equipment.
+1 Seconding this. My limited experience with pandas had a non-trivial number of moments "?? Is it really like this? Nah - I'm mistaken for sure, this can not be, no one would do something insane like that". And yet and yet... Fwiw since I've found that numpy is a must (ofc), but pandas is mostly optional. So I stick to numpy for my writing, and keep pandas read only. (just execute someone else's)
+1. I also use Samsung Galaxy Watch (Active2) and find it to work well for me. I'm surprised more people don't know about it. I know couple of people that suffer from high BP. They showed little interest when I told them about the watch.
Regular external calibration helped me build confidence. I was suspicions if the watch worked, and how well it worked. Calibration is every month - it expires. I take 3 measurements in parallel watch-cuff, and tell the watch App what the cuff measured. I didn't mind it initially. I now find it annoying tbh. I see why Apple would like to get rid of it. Post calibration, I measure 3 times in parallel both watch and cuff to check if they match. If the difference is within 5% - then fine. In 3 years it's happened twice where they differed more and I had to recalibrate. From recollection - both times with a reason, a TV was turned on in living room.
I have the impression that BP measurements differ though the day (lower in the morning, higher in the evening), depending on activity or at rest of course, I think there maybe some seasonality too (winter v.s. summer). Measuring BP when going to the doctor once a year strikes me as a wholly inadequate. If the BP is absurdly high out of normal range - then yeah, we learn something is wrong with the patient. But in any other case - don't see what can be deduced from 1 random measurement.
To check if the calibration is important, I had my wife use the watch, that is calibrated on me. She got absurd readings. So yes - the calibration is important and is tied to the person that did the calibration. (edited to add this)
Most people don’t know about this feature on the galaxy watch because it’s only approved in certain countries, though you can hack the SHM to enable it.
Not the OP, but I remember reading about many twists and turns on the road to various inventions described in Matt Ridley's "How Innovation Works". I personally like "Happy Accidents. Serendipity in Major Medical Breakthroughs in the Twentieth Century" by Morton Meyers.
You have agency. Yes - the system provides incentives. However, you are not some pass-through nothingness to just accept any incentives. You can chose to not accept the incentives. You can leave the system. You're lucky - it's not a totalitarian system. There will be another area of life and work where the incentives align with your personal morals.
Once you bend your spine and kneel to bad incentives - you can never walk completely upright again. You may think and convince yourself that you can stay in the system with bad incentives, play the game, but still somehow you the player remain platonically unaffected. This is a delusion, and at some level you know it too.
Who knows? If everyone left the system with bad incentives, it maybe that the bad system collapses even. It's a problem of collective action. The chances are against a collapse, that it will continue to go on for some time. So don't count on collapse. And even if one was to happen in your time, it will be scorched earth post collapse for some time. Think as an individual - it's best to leave if you possibly can.
The list of things the consider is long almost exhaustive. Most of the points have already been made, some by other people too, if you follow these things online. While there is nothing there earth-shattering, it's good to see it all nicely laid out. People dismissing everything with a sentence, are not doing it justice.
However. The report is thin on recommendations "What is to be done now?", or did I miss them really!? That really surprises me. If I were to define it in two sentences: UK central gov is ultra centralised in Whitehall (#1) (almost Stalinist in that respect!!), and then that centralised power is further held in a strangle hold by the Treasury Brain - Deep State - OBR complex (#2). So together they are like Centralisation^2.
Atm that ensures - #1 nothing can be done by anyone else, and #2 they've settled on "do nothing" b/c everything costs money, and #2 care disproportionally more about that cost, and much less about what capability assets etc that cost buys UK. Once nothing tangible is done, laziness is naturally encouraged, activity punished, and every promotion every jostling in the hierarchy becomes entirely a beauty contest, divorced from any real-word results. All this seen afar by outsider without any insider knowledge. Just listening to Rory Stewart and Dominic Cummings relying info about the inner workings of UK gov machine.
So any intervention would have to come (walking backwards) into point #2, and or at point #1.
Point #2 means smashing the power of the Treasury. They are good servant, but a bad master. No.10 (the PM) needs to re-assert itself over and above No.11 (the FinMin) by 1st-ly taking decision powers out of No.11 hands. No.10 would have to work much more on getting things done, take interest in that, and leave behind the media PR Circus it's totally devoted to atm. Just assume you will serve 1 term (or less!), drop the "permanent electioneering mode" (that nowadays the position goes into while in gov??), don't think of re-election and try get stuff done, assuming you've run out of time. Can't see how a change like that can come but directly from the top, from the PM. Supported by legislation and votes to dismantle the 1000 roadblocks, via the political power of MPs voting in Parliament.
Point #1 means decentralisation of power. UK for a long time used to have vigorous local authorities that could innovate in various ways. Admired once at the riches of local gov building in Birmingham (UK 2nd largest city). The idea that some pimply fresh graduate 20yo SpAd 120 miles away in London would be telling these people how many pencils they should buy I presume would have seemed to the people running the City gov absurd, like a Monty Python sketch to us.
Mid-term UK needs to build another 10mil strong cities agglomeration, e.g. Liverpool - Manchester - Leeds - Sheffield - Hull, by connecting them with transport links, and building housing along the links for commuting whichever way people fancy. With Birmingham half-way between this new MegaCity and the old London. So London gets some competition, and raises its game too. Atm feels London is coasting because TINA.
The literal truth can hardly be a "propaganda line". More constructive criticisms are 1) private sector could have done better with the same finance (ultimately - resources); and 2) interest on that is paying basic income to people that already have money. I am sympathetic to both. So - the debt itself is a signal, a symptom. The solution is not to treat the symptom, but the causes. On 1) - you can say the state needs to do less. Just go down the list of things the state does, and propose those be reduced. From that PoV, the federal gov being "insurance company with and army" is basically correct. So to reduce gov spending, need to look at those two top of the list items. On 2) - that's easier imo. The IR rate is a decision, a vote, by a monopolist, controlled ultimately by the legislature. Just reduce the IR if that's an issue. Then fight inflation (if needed) with other means. People have forgotten that - but can relearn quickly.
You are entirely right.
If the same number was called "private credit" I believe no one would bat an eyelid about it.
Of all the non-issues to spend time on - this is the least worth anyone's time.
You are right and IDK why you are downvoted. Few units of perceptrons, few nodes in a decision tree, few of anything - they are "interpretable". Billions of the sames - are not interpretable any more. This b/c our understanding of "interpretable" is "an array of symbols that can fit a page or a white board". But there is no reason to think that all the rules of our world would be such that they can be expressed that way. Some maybe, others maybe not. Interpretable is another platitudinous term that seems appealing at 1st sight, only to be found to not be that great after all. We humans are not interpretable, we can't explain how we come up with the actions we take, yet we don't say "now don't move, do nothing, until you are interpretable". So - much ado about little.
"I don't believe the 2nd law of thermodynamics. (The most uplifting video I'll ever make.)"
https://m.youtube.com/watch?v=89Mq6gmPo0s
I come to entropy from Machine Learning, Information Theory and probability. For me it's fairly straightforwad. Of interest and useful - but nothing mysterious there.
The p.d.f. aka a fancy histogram is my current best knowledge of how many times an outcome is expected to happen. When I do one experiment - I can't tell the outcome. I can only count the number of different outcomes, without being certain when will any of them appear exactly.
A flat p.d.f. means my knowledge is poor: every outcome is about equaly possible. I'm very ignorant (high entropy). A spiky p.d.f. means I have good knowledge: some outcome is much more likely (low entropy). In extremis the p.d.f. is a Dirac impulse - that's deterministic knowledge.
The only mildly interesting thing is when a new observation reduces my knowledge. Say right now I'm fairly certain I have not got cancer. My chances are 90:10 for my age. Tomorrow I take a test, the test comes back positive. Of people of my age that test positive, aboout half have cancer for real. After the test my chances are 50:50. Now I am perfectly ignorant whether I have cancer or not. Whereas before I took the test and got a positive result, I was very certain I have not got cancer. The new information (positive test result), transformed my probability of cancer from spiky marginal P_Y(y)={0.9,0.1} to perfectly ignorant conditional P_Y(y|+ve test)={0.5,0.5}.
This example is from "How to measure the information gained from one symbol" by DeWeese and Meister (https://pubmed.ncbi.nlm.nih.gov/10695762/).