Wednesday, October 5, 2016

Midweek Nerd Pride Post

The thunder down under [brought down the electrical grid]

The South Australia state of, well, Australia serves as a reminder that correlated risks aren't just for financial crises:

Australia: proof the correlated risks aren't just for finance

Source for the timeline and quote: "PRELIMINARY REPORT – BLACK SYSTEM EVENT IN SOUTH AUSTRALIA ON 28 SEPTEMBER 2016" by AEMO  (PDF file)

(My previous analysis, from preliminary and incomplete data, suffered from lack of granularity.)

The memes started immediately:

I don't think the australians deserve this meme, though eight people have sent it to me so far, so it's clearly popular. The problem isn't the wind generation per se, it's the lack of large scale batteries, as explained here by Donald Sadoway of MIT.

Video killed the radio star bad science popularizer

I wrote a post on Sunday that was a little negative on popular science popularizers, but forgot to mention that the emergence of alternative media platforms, like YouTube, could counter the decline in the quality of science popularization on mass media. So here are a few examples from this week.

There's always something fun to watch on the Numberphile YouTube channel:

I watched the Norwegian TV show "The Heavy Water War" on Netflix in the last three evenings (two episodes per evening, not really a 'marathon') and therefore rewatched these fun videos from the YouTube Periodic Videos channel:

Walter Lewin shows how to make teaching come alive, teaching optics to high-school science teachers. He has the advantage of teaching Physics, with practical demonstrations (that is, using real physical objects), unlike those of us who teach more indirect material (like, say, decision-making or strategy or analytics):

As a contrasting example (of how not to talk about Physics), here's the announcement for the 2016 Nobel Prize in Physics:

I'm sure there are pretty good Physics instructors in Sweden. But instead they did the European bureaucratic thing and had a member of the committee give the talk, apparently without any sense of how to weave a narrative or what to emphasize. Yes, I understand that he's a physicist, but not one that can present physics to a regular audience. (Compare with Walter Lewin above.)

As a bonus, the announcement also showcases the total lack of interest in science of the reporters assigned to cover this scientific prize. But that's par for what passes for journalism these days. Why bother with knowledge when you can ask "human interest" stories?

Staying with Physics, here's a fun lecture that basically tells us how much we don't know about the universe. Really, we know very little:

And, given the section title, here are the Buggles singing "Video killed the radio star"

Still geeky, but more technical…

Stats… Yesterday was the official October BARUG (Bay Area R User Group) MeetUp at Santa Clara University, so I went early and enjoyed their beautiful campus:

At Santa Clara University for the October BARUG meetup.

There was a talk about Benford's law (a statistical regularity of the distribution of the first digit of large sets of numbers. It was inspired by this Business Insider piece.

Two presentations showed political forecasting; not my thing, but one was an interesting application of kernel methods, particularly KRLS (kernel regularized least squares): Slides on GitHub here. Definitely worth checking out.

More stats… Nassim Nicholas Taleb (backed by Pasqualle Cirillo) is locked in a deathmatch with Steven Pinker and some of his more statistically-proficient supporters. Alas, I think this is game, set, and match for Taleb:

Even more stats… Andrew Gelman finds closure in his argument with Amy Cuddy and co-authors, from an unexpected side: one of the co-authors has now all but retracted the paper.

Some commenters raise an interesting question, for which I have no answer: given that the work on which hiring, promotion, and tenure of these authors was base is at best incompetent and at worst fraudulent, should these authors resign (or their institutions fire them)?

On the one hand, it might seem reasonable that when the work that supported some institutional decisions (hiring, promoting, tenure) is shown to have been invalid, those decisions should be revisited. There are legal questions that make this point moot, but in principle it appears to be a reasonable idea.

On the other hand, any such actions may make it less likely that people might own up to problems with their papers in the future. This isn't a small thing, as some of these errors might be" par for the field" at the time of writing but unacceptable a few years later.

In a follow up post, Andrew asks "Why is the scientific replication crisis centered on Psychology?" Good read, but here's a thought: When a paper "fails to replicate," the problem isn't in the replication, it's in the original research.

Linkage galore

Nobel Prize in Chemistry 2016 awarded to three pioneers of molecular machines via Kurzweil Accelerating Intelligence.

Designer Babies, by Peter Diamandis.

Bacterial molecule trains the immune system to tolerate infection without inducing illness, on Medical Xpress, via

California eyes unusual power source: its gridlocked roads via No, not Solar Roadways nonsense; piezoelectric generation. Still probably nonsense in any appreciable scale, though.

This Huge Robot Will Drive Up and Build You a House, via Singularity Hub.

Privately-developed rocket aces abort test, via Spaceflight Now. Blue Origin's puddle jumper passed its abort test, so it's one step close to being able to take passengers to "space."

A book that's going on my reading list: The Paradox of Stupidity
As Jan Wallander, the ex-chairman of Sweden’s Handelsbanken, said: ‘Business leaders are just as fashion-conscious as teenage girls choosing jeans.’ Many companies adopt the latest management fads, no matter how unsuitable they are. If Google is doing it, then it’s good enough reason to introduce nearly any practice, from mindfulness to big-data analytics. - You Don't Have To Be Stupid To Work Here, But It Helps.