jarvist

Physics is really a working knowledge, like computer programming. You have to be able to solve problems. All the 'practical' and 'intuitive' aspects of theoretical physics are built by working on problem sheets. A lot of progress in understanding (both personally and for the field!) is in tackling apparent paradoxes.

Susskind's courses are very much overviews / appreciations. For each of these areas (GR, StatMech, Quantum etc.) you would expect several 30-hr lecturer courses + problem sheets (or equivalent working through a textbook) to gain a deep knowledge.

Could you point me in the direction of a single YouTube channel that goes into more depth than a Masters program?

Other than occasional captured lectures, everything 'educational' on YouTube seems to me to be optimising for that TED-style sense of wonder / happy familiarisation with the content, which is poorly associated with learning outcomes.

The issue that I find with Google scholar is that it seems very biased by my publications. So it's very good at telling me the latest work related to what I used to work on, but not necessarily what I am currently interested in.

There is something general: https://arxivist.com/

Interface is much less polished than ArXiv-sanity. It can also send a daily email of the top 5 for you, which I find very useful.

The Raspberry Pi 400 is basically this. https://www.raspberrypi.org/products/raspberry-pi-400/

CEO of AstraZeneca: "But the UK contract was signed three months before the European vaccine deal." "When we entered the agreement with Oxford, they had already been working with the UK government on this. "

https://www.repubblica.it/cronaca/2021/01/26/news/interview_...

As you already have the gnuplot commands, perhaps using Julia + https://github.com/gcalderone/Gnuplot.jl might work? It's a very lightweight package, so even using it directly should be OK.

I do all my plotting with Gnuplot.jl, as gnuplot is fast, and I can save .gpt files which reproduce the plots for later reference and making publication-quality.

Sort of... but the wind turbines often don't point well when you're at sea, and you intentionally anchor in calm places to avoid the swell. Sailing specific PV would often have 40 silicon cells rather than 36 in series, to provide some headroom for partial shading. Watt & sea hydro-generators have become quite ubiquitous recently, as if you're making passage your entire vessel is a machine for concentrating wind energy: https://www.wattandsea.com/index.php/en/products/cruising-hy...

From the brochure, it's 2300 tons of steel in the substructure vs. 670 ton weight of the towers. It's not stated what fraction of the ballast is fixed, and what fraction is water. Normally the fixed ballast would be rock. I don't know how these numbers compare to a shallow water fixed installation, or even an onshore turbine (they must have some foundations!).

The new LUMI EU supercomputer in Finland will put its waste heat into a district heating system: https://www.hpcwire.com/off-the-wire/csc-lumi-supercomputer-...

I'm sure smaller scale data centre heat to district heating schemes must already be in place. Fundamentally you are using the same technology to cool the data centre (a heat pump), just pushing that heat into hot-water / steam, rather than dumping into the air.

For matrix manipulation based Fortran scientific codes, ifort/MKL can give +30% compared to gfortran. It's difficult to disentangle where the speedup comes from, but certainly as jedbrown aludes to, the Intel compilers seem to make a better go of poorly optimised / badly written code.

For C based software, its a much closer run thing, and often sticking with GCC avoids weird segfaults when mixing Intel and GCC-compiled Linux libraries.

Thank you! I wasn't aware of this. But this is only a replacement for libm (i.e. basic trig, exp function), not the matrix-orientated BLAS, LAPACK and SCALAPACK routines that scientific codes spend >90% of their time.

Most high performance software on super computers uses the Intel C and Fortran compilers, and much engineering and scientific software on workstations uses the Intel Maths Kernel Library (MKL) for high performance linear algebra.

Now that AMD EPYC processors are powering a lot of next generation super-computer clusters, we're going to have to figure out some workarounds!

> But nuclear power plants? Completely uncorrelated. If each one produces power 85% of the time, then you need two, and you'll almost always have power.

Not really true. French nuclear output drops every time there's a European heatwave. It used to be the only time that the cross-channel link rank backwards (UK powering France).

I think France responded to this in the context of climate change by changing the protection standards for their rivers?

https://www.theguardian.com/world/2003/aug/12/france.nuclear

Starting with the 2010 introduction of the A* grade, the proportion of grades awarded has been almost constant (25% A/A*, 25% B, 25% C). This is in contrast to the consistent 'grade inflation' visible for the previous twenty years. https://en.wikipedia.org/wiki/Grade_inflation#UK_A-Level_cla...

Previous YoY was -1.9%/-1.5%.

How could +5% possibly be an actual trend? What new magic teaching innovation have independent schools done for this year's cohort, that wasn't available in previous years or to teachers in the state sector?

Table 9-10: https://assets.publishing.service.gov.uk/government/uploads/...

No, teachers submitted both predicted grades, and in large cohort classes (such as English and Maths), a rank ordering of the entire cohort.

It's more complicated than this. They haven't just taken the rank ordering by the teachers and projected onto a curve.

Here is the 319 page document published on results day. This is the totality of the current description of the 'algorithm'. https://assets.publishing.service.gov.uk/government/uploads/...

There's some kind of predictive linear model (page 40) which is based on a student's prior attainment (i.e. GCSE and SATs 'data'), but then modelled at the level of the exam centre (i.e. school). These models are separately fitted for different bins (I think ten are shown at one point) of achievement.

The whole document is pretty incoherent. I suspect the 'algorithm' is a set of excel spreadsheets.

Almost all scientific computing. In spite of considerable effort, there are relatively few workloads that use GPUs well, and even for codes which do work on GPU, you tend to end up with partial functionality, and a very restrictive region of parameter space where they work efficiently. (And often you run out of memory very quickly - scientific workloads typically run multi-node and share RAM across the nodes via Global Arrays or equivalent. Using multiple terabytes of memory is absolutely normal.) The computation in deep learning is actually quite simple (matrix-vector operations), and still it has taken considerable effort to develop efficient kernels (such as cuDNN).

The next UK supercomputer is based around AMD's 64 CPU chips for this reason. One concern is that we may now be limited by the shared memory bandwidth. https://www.nextplatform.com/2019/10/18/amd-cpus-will-power-...

The vast majority of time on research supercomputers is used in running Fortran codes. (Click on 'languages' https://www.archer.ac.uk/status/codes/ , this is the UK national facility.)

And yes, as you've noticed almost all linear algebra is written in Fortran (BLAS, LAPACK, ScaLAPACK), so no matter what language you've written something to do some useful computation in, it's likely that most of the cycles are spent in Fortran!

I have these shell aliases configured, to check and delete the master connections. (Often I end up with stale connections when I have a VPN up, and then just sleep my laptop.)

alias ssh-MasterConnection-check="ssh -O check " alias ssh-MasterConnection-exit="ssh -O exit "

+ then you use them as 'ssh-MasterConnection-check host' etc.

grep -B / -A / -C [lines Before / lines After / lines of Context] is very useful for when you have multi-line text you want to extract.

I read this as a kid: https://www.amazon.co.uk/Cartoon-Guide-Genetics/dp/006273099... It provides a very useful bridge from a mechanistic / physic-y / computer point of view, to the more messy biochemistry. Still, whenever I think of a biochemical process, it's a cartoon from here that comes to my mind's eye.

I'm not sure if this article hammers home the key point: extracting carbon from the air always takes more energy that was gained by burning the fossil fuel to make that carbon. This is inevitable due to the entropy cost of de-mixing CO2 from the air, and from the limited efficiencies of thermal power generation (~33-40% for a coal power plant, perhaps 20% in an internal combustion engine), and limited efficiency of the carbon capture technology.

So it only makes sense to spend energy on sucking CO2 out of the atmosphere, once you've already de-carbonised everything which you possibly can.

Maybe it has a future role (>2050? >2100?) to start using excess low-carbon energy to take CO2 levels back to their pre-industrial level, or to constantly offset hard-to-decarbonise technologies such as air travel.

By vibrational spectroscopy. You compare the microwave / infrared spectrum of a compound measured in the lab, to data from telescopes.

Fullerenes are by far the largest molecules positively identified in space. Due to the high symmetry of the molecules, they have a very specific and quite simple vibrational fingerprint.

https://www.nature.com/articles/369296a0 https://en.wikipedia.org/wiki/List_of_interstellar_and_circu...

No, it sounds like a telescope focusing onto a single pixel (i.e. just the back of the curtains), and then using the output from this pixel with a lock-in amplifier tied to the broadcast signal. The sensitivity would be enormous due to the lock-in.

With a CRT you would be able to do this with <microsecond resolution as the scan lines went across.

In today's LCD world, I guess something similar could be done for a whole pixel-average of a whole frame, and then look at the time-series of that to correct for lag in digital television / internet etc..

Like Chernobyl, AGR reactors do not have secondary containment: the idea was that as they are gas cooled (no water to phase change) and have enormous cores (so much more time to deal with heat build up), they are safe enough to not need it.

In terms of commercially available bulbs, the 'Philips Master LED ExpertColor' (all in spotlight 'GU10' etc. format) have an extremely high quality halogen-like spectrum.

Chapters 0 and 1 in this book by Mattuck are also a fantastic read for a general audience interested in what's going on 'under the hood'. https://www.amazon.co.uk/Feynman-Diagrams-Many-body-Problem-...

Feynman's own book 'QED: the strange theory of light and matter' is another good read, but strangely it is a bit lacking in illustrations. (And the illustrations that are there are a bit dry compared to Mattuck!)

If you want to be more environmental, use power when the current carbon intensity of the grid is low. Effectively you are then providing demand-side management to reduce overall carbon dioxide production. https://carbonintensity.org.uk/