"Thats because running Stanford Universitys gamut of academic obstacles requires a mechanical mindset. For example: Last month, in the minutes leading up to a three-hour computer science exam, I made small talk with a fellow grad student. He said: I hope they dont make us think. My machine brain understood. He meant: I hope we only have to do math. Calculations are quick, but it takes time to think through the implications of our decisions. Speed and efficiency are higher priorities."
The writer is "a doctoral student in education data science" at Stanford.
I have an old book "Up The Infinite Corrodor", about "Learning to think at MIT."
On a sunny day (Tue, 13 Dec 2022 08:47:51 -0800) it happened John Larkin snipped-for-privacy@highlandSNIPMEtechnology.com wrote in snipped-for-privacy@4ax.com:
I try to avoid math, neural nets do not use math come from a family that had watch makers.... Math is an aid, not a solution. Most 'equations' and formulas are just an incomplete description of reality, divide by zero is all over modern 'science'
Not so. Just as logic has a mathematical-logic extension, so neural nets have a kind of multivariate correlation going on; it's useful to mathematically model all sorts of systems, and 'use math' isn't going away, because the mathematical insights are... useful in deep ways.
yeah, because description in natural language has proven impossible too many times, and the mathematical notations are sometimes the best description
For good cause; we see a problem and we keep trying until we solve it; division by zero indicates it isn't time to stop trying, yet. Modern science isn't going to move forward until each new (divide-by-zero or other) challenge is mastered.
On a sunny day (Wed, 14 Dec 2022 01:26:10 -0800 (PST)) it happened whit3rd snipped-for-privacy@gmail.com wrote in snipped-for-privacy@googlegroups.com:
Typical example is a the black hole 'singularity' No clue what the mechanisms is, divide by zero gives zero radius for infinite mass. So endless crap about singularities. My rule: Something will always break down in nature, so singularities cannot exists. And the Einsteinian dogma and field theory. We need a MECHANISM, a Le Sage type particle does away with all that bull. From a mechanism you can predict things, from math only in the best case only quantize things, that is ALL math is good for.
No, neural nets do not use math But you can use math to build one. Well the ones I wrote did not use a lot of math IIRC use math to describe it OK.
Not a day goes bye without AI (neural nets) coming up with solutions to for example medicines. Usually by just trying endless combinations end evaluating the best results to do new combinations. Hey when I started coding a neural net in the eighties I think it was, using an example from a project from a German professor, its behavior was what interested me, behaved just like a humming bean (human being if you will....) !!
The trend is universal in academia. Mandatory pages of equations everywhere. It's especially bad in engineering where the theses and papers are generally useless.
The interesting thing about the quote above is the concept that math is easy, solving equations is a thought-free mechanical process, and thinking is too much work.
Oxford is pretty but the food is better in Palo Alto.
Elseiver is a big part of the equation-packed paper mill. Spending one's 20's and maybe 30's authoring peer-reviewable academic papers is not the best path to designing electronics.
Equations analyze, but Spice can do that, and there are even software equation solvers. I hope we only have to do math.
Or perhaps John Larkin isn't good enough at mathematics to understand what they are saying.
Most academic papers are un-inspired hack work, but then so is most electronic design
Solving some equations can be automated. Math-heavy courses tend to be a weak on connecting the neat equations to the messy reality that they are supposed to represent.
Depends where you eat.
formatting link
is famous. Heston Blumenthal goes in for some gimmicky tricks, but the results are generally held to be good.
If you want to publish a paper in peer-reviewed journal it goes through peer-review. Peer-review doesn't get rid of all of the rubbish, but it does thin it out. "Peer reviewable academic papers" aren't any kind of separate category of academic paper - if it's an academic paper it has to go though peer-review.
The kind of academic that churns out academic papers on circuits that nobody is ever going to put into production is a waste of space, but there aren't all that many of them.
Equations don't analyse anything. Neither does Spice. Both are tools that some people use to work out what is going on. John Larkin doesn't seem to be one of them.
On a sunny day (Wed, 14 Dec 2022 06:59:27 -0800 (PST)) it happened Fred Bloggs snipped-for-privacy@gmail.com wrote in snipped-for-privacy@googlegroups.com:
Bull electronics or electricity without electrons leads to nonsense When it was found that current could flow in a vacuum tube electrons was the MECHANISM that explained that Einsteinian crap field theories will ALWAYS fail as it provides no mechanism as does Newtonian gravity as we now see the problem were we look for dark matter. Experiment comes first, then mamaticians make a model that sort of describes a relationship between observed variables, the mamatician then claims he invented it and makes predictions that fail 200% as in the no current can flow in a vacuum tube case and singularities and a million other cases. They FAIL to see their formulas are a INCOMPLETE description of reality, and in pushing their silly ideas on humanity make a mess like in epicycles where math became so complicated to derive the motion of the planets that only a few selected mamaticians could understand it and supported by religious powers stopped advancement or even killed advancement for many hundreds of years. Now we see silly people using 'spices' not where it is intended: food, but to design 'tronix while if you look at the POV from the electron things are really very simple.'Respect the electron!!!
As a prophecy this hasn't got much traction. Einstein did explain the precession of the orbit of Mercury - the mechanism is the distortion of space-time, which Jan doesn't seem to be able to grasp. Gravity waves are additional supporting evidence.
Except the if you do proper general relativity based simulations of our galaxy you don't seem to need dark matter, though you do need big computers to do it.
Dirac predicted the existence of the positron in 1928, which prompted the search that found it in 1932. Zwicky invented dark matter at much the same time and we still haven't found anything that would do the job.
None of which exist outside of Jan's deluded brain.
Everybody knows that formulas are always incomplete. Most people understand that they are frequently quite good enough.
Epicycles sort of worked. Ellipses worked better. Newton provided a justification for ellipses, but couldn't explain the precesion of the orbit of Mercury, which Einstein could.
It's respectable in vacuum - less useful in condensed matter where electrons can act in combination.
Not only that but he predicted that the speed of light in a vacuum would be a fundamental constant on nature way before Einstein took it as an axiom and ran with it.
Gauss was truly a genius. I'm reading one of his early works now (in translation) and it includes computors tricks to get a solution in the days when everything was computed manually. One sneaky one is to combine things in a ratio 1/10 (which is a simple shift in decimal digits).
Physicists are well aware of the limitations of their mathematical formulas. What you need to understand is that mathematics is much less ambiguous than hand waving just so stories that you and Larkin prefer.
Mathematics in its purest form provides proof of correctness for both hardware and software designs. Way better than "gut feel".
The solar system is a complex linked dynamical system.
The modern theory is as simple as it can be made and still get the right answers. Until comparatively recently amateur telescopes could reliably "Goto" any planet apart from the moon. The theory of lunar motion is so horribly complicated that the formulas within the capabilities of early computers couldn't hit the moon reliably at all points in its orbit.
Ultimately it *has* to agree with reality or it is no use at all.
Spice is a remarkably good electronics simulator.
For every complex problem there is a simple wrong answer.
Equations concisely describe the relationships of the variables - how else would you seek to describe and solve an engineering problem.
Make a model of it in play-dough? Music or poetry?
The maths I did at university certainly wasn't easy even if you were gifted. The first week of the long vacation course was designed to weed out anyone who might not make the grade (about 60% drop out rate).
Cambridge is prettier.
I have returned to astrophysics during lockdown having discovered that one of my little computational tricks is better than state of the art. The hardest part has been getting to grips with Latex typesetting and Gnuplot for publication quality graphs. There are plenty of equations in the paper but only one of them, the true novelty really matters. The rest of it is all a scaffold to make it useful to other practitioners.
It is a curious combination of a very old manual computors trick and a modern Pade approximation for sin leading to a cubic equation that solves a transcendental problem fully accurate to the quintic form.
Mathematica, Maxima, Reduce, Camal and the like are only as good as the question that you ask of them. They do most of the donkey work of differentiating and substituting equations with no risk of error.
The next generation of AI guided tools will show true creativity just like a human. Mathematics is mostly pattern matching and recognising the intrinsic symmetries of a problem and then finding a way to exploit it.
The computer tools give us more leverage and allow brute force attacks that a human could not contemplate and would inevitably make a mistake at some point. Poor old Shanks pi calculation for example:
Physical models are often pretty useful, actually. I spent a pleasant afternoon once, making a 10,000x scale model of an optical antenna modulator/detector gizmo that showed a 95% coupling efficiency and 20-dB or so on/off ratio. I then spent the next four years trying to do the same thing on-chip.
Yeah, the idea that doing math is thought-free is just cracked. Mathematical notation is a technology of thought--one that allows us to make correct inferences of a power and complexity completely inaccessible to ordinary discursive reason. (A good notation is also highly suggestive of avenues to explore.)
It has to be guided, of course, like any other technology.
From my reading, modern papers tend to be less mathematical, and rely more on some crufty bit of grad-student code that they won't show you because they're too ashamed of it--but really, it's all OK: "trust us, it's true, it's true." (Niall Ferguson, I'm looking in your direction.)
It's far more useful to have the opportunity to go through the math yourself and _show_ if the paper's claims hold up. Nowadays with Maxima or Mathematica or <insert your favorite symbolic math package here>, checking the math is much easier. It's still a serious time investment, of course, so most folks don't do it that often.
Those things are almost worthless unless you're aiming to have your sheepskin do the talking for you. In choosing a school, it's far more important to figure out who you want to work with. There's lots of dead wood everywhere--for instance political animals, hacks, and ass-kissers.
I did my doctorate at Stanford, which was fine, but the key thing was working at the old E. L. Ginzton Laboratory. Just about everybody who's anybody in optical physics has a Ginzton connection--prof, student, visiting faculty, or something. Dunno how that's held up since they knocked down the old building in 2010, but it was certainly true then.
That was a case of blind luck and bloody ignorance on my part--I didn't know any of the folks I wound up working with.
Yup. I've only submitted one pure theory paper, when I was a student--I had to withdraw it from the review process when I found out that the main result had been published by Lord Rayleigh. One of my edgier colleagues(*) asked me, "So, Phil, how does it feel to be at the forefront of 19th Century science?" ;)
Cheers
Phil Hobbs
(*) Dr. John Fox of SLAC, who later invented a beam-cooling system for particle accelerators that AFAIK is used absolutely everywhere.
Spice. Breadboards. Consultants. Nonlinear calculus is tough.
Equations are fine, as long as they don't make one assume that they are always what's actually going on.
Some, very few, people can get creative insights from equations, but to most people they are just things to mechanically solve without thinkling, like the Stanford kids.
Why does "a doctoral student in education data science" need a lot of equations?
On a sunny day (Thu, 15 Dec 2022 17:22:02 +0000) it happened Martin Brown <'''newspam'''@nonad.co.uk> wrote in <tnfl44$mg$ snipped-for-privacy@gioia.aioe.org>:
University mental health services at Stanford is probably a lot like mental health services at most universities in the US, so long as you don't need anything they'll always be there for you.
On a sunny day (Fri, 16 Dec 2022 02:18:10 -0800 (PST)) it happened Anthony William Sloman snipped-for-privacy@ieee.org wrote in snipped-for-privacy@googlegroups.com:
Well, my stuff works, you have nothing to show apart from Baxandal that was not from you anyways,
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.