Later On

A blog written for those whose interests more or less match mine.

Archive for the ‘Software’ Category

Coarse-brush week, and today is Leviathan and the Baili 171

with one comment

Yesterday I used my Omega 20102 and today you see the Omega Mighty Midget, a mix of boar and badger. I do wet the knot and let the brush sit while I shower to soften the boar bristles. A reader commented on his fondness for horsehair, and those brushes too have a pleasantly coarse feel on the face — not rough, but with a perceptible grain. So I thought II’d go through some of the coarser brushes in my collection for a pleasant change of pace.

The Mighty Midget, though, really doesn’t feel all that coarse. The badger smooths it out quite a bit. It did make a mighty fine lather from one of my favorite soaps, this one from Barrister & Mann.

The Baili 171 is a remarkable razor: $6 at the link (and I have no affiliation with the company — I’m just a customer), and it shaves like a dream. It’s so comfortable it doesn’t feel as though it’s doing much, but the result today is as smooth as one could want. I also like the looks and feel in the hand. It’s somewhat unusual in that it secures blade alignment through corner brackets instead of the usual studs from the cap (or baseplate). Works like a charm.

A good splash of Leviathan aftershave — I love the fragrance — and I’m set for the day, which will include some afternoon chess. I downloaded a free (and quite nice) chess-clock app for my iPhone, one provided by I recommend it if you play any two-person strategy games (chess, Go, checkers, or the like) since it ensures that the games move along, plus it’s easy to give (or receive) a time handicap — e.g., the stronger player gets 10 minutes and the weaker player gets 20. It’s not so cut and dried as that seems, since obviously the strong player will be thinking while the weaker player’s clock runs, but it can help — particularly if the division is 5 minutes vs. 25 minutes.

Written by LeisureGuy

30 June 2020 at 10:33 am

Posted in Chess, Games, Shaving, Software

Algorithm-governed interactions are often convenient, sometimes enraging, and occasionally dangerous

leave a comment »

Here’s an example of the enraging sort. The comments on YouTube for this video are interesting:

Written by LeisureGuy

29 June 2020 at 9:33 am

‘A chain of stupidity’: the Skripal case and the decline of Russia’s spy agencies

leave a comment »

Luke Harding reports in the Guardian:

In 2011 I was in Libya reporting on the civil war. Rebels backed by the US, the UK and France were advancing on the capital, Tripoli. The insurgents moved forward through bombed-out towns as Muammar Gaddafi’s forces retreated. Coastal cities in the west and east, oil refineries, Roman ruins and temples – all fell, one by one, as the regime lost ground.

These were dangerous times. In the town of Zawiyah I found locals celebrating victory in the main square. They were shooting in the air and doing wheelspins and skids in their cars and trucks. Gaddafi’s soldiers had left the previous night, fleeing down the road. I saw a small boy, maybe eight years old, stomping on a Gaddafi flag. “The city is ruined. No problem – we will rebuild it,” one local, Tariq Sadiq, told me.

The signs of battle were everywhere. The square’s four-star Zawiyah Jewel Hotel was a ruin. The lobby was filled with rubble. Mattresses where Gaddafi’s soldiers had slept lay strewn among crates containing mortar cases and empty plastic water bottles. The air crackled with jubilant gunfire.

The celebrations turned out to be premature. From their new positions, and without warning, Gaddafi’s army began shelling the square. I took shelter indoors. First one mortar, then six more. Each was a loud thunderclap, a sudden affirmative whomping, followed by puffs of black smoke.

At that moment, I wasn’t much interested in the types of munitions that were raining down. There was a simple urge: to escape. My role, as I saw it, was to tell the stories of those unwittingly caught up in conflict. I had brought to Libya the usual tools of a frontline correspondent: flak jacket, satellite phone and first-aid kit, carried in a rucksack.

A man named Eliot Higgins was following events in Libya, too – not from the front line, but from his home in the east Midlands. Specifically, from his sofa. It was a safer place to be – and, as it turned out, as good a perch as any from which to analyse the conflict, and to consider questions that, in the heat of battle, were interesting, but seemingly unanswerable. Questions such as: where did the rebels get their arms?

Higgins recalls growing up as a shy “nerd”. According to his brother Ross, Higgins was an obsessive gamer and early computer enthusiast. He liked Lego, played Pong on an antediluvian 1980s Atari and was a fan of Dungeons and Dragons. He spent hours immersed in the online roleplay game World of Warcraft, where participants pooled skills and collaborated across virtual borders. His instincts were completist: he wanted to finish and win the game. This would prove useful later on.

Higgins tried for a career in journalism and enrolled on a media studies course in Southampton. It didn’t work out, and he left without a degree. Next, he earned a living via a series of unlikely administrative jobs. One day Higgins logged on to the Guardian’s Middle East live blog. Libya was the centre of international attention. Higgins made his own contributions to the comment section of the Guardian blog, using the name Brown Moses – taken from a Frank Zappa song. The blog often featured videos uploaded by anti-regime fighters. There was fierce debate as to whether these images were authentic or bogus.

One such video showed a newly captured town. The rebels claimed it was Tiji, a sleepy settlement with a barracks that had been recently bombed by Nato jets, close to the border with Tunisia, and on the strategic main road leading to Tripoli. There was a mosque, a white road and a few little buildings with trees around them. The video showed a rebel-driven tank rolling noisily down a two-lane highway. There were utility poles.

Higgins used satellite images to see if he could identify the settlement and thereby win the discussion. The features were sufficiently distinctive for him to be able to prove he was correct: the town was Tiji. “I’m very argumentative,” he says. It was the first time he had used geolocation tools. He realised he could collect user-generated videos and later work out exactly where they had been filmed.

Shortly afterwards his first child was born. Higgins combined his new childcare duties with online research. Meanwhile, the uprisings in the Arab world spread. Soon Syria was at war, too.

What began as a way of scoring points over online adversaries evolved into something bigger. Smartphones with cameras, social media, Facebook, Twitter, Google Earth, Google street view, YouTube – the digital world was multiplying at an astonishing rate. This stuff was open-source: anyone could access it. By cross-checking video footage with existing photos and Google maps, it was possible to investigate what was going on in a faraway war zone.

These techniques offered interesting possibilities. Open-source journalism might be applied to the realm of justice and accountability. Sometimes soldiers filmed their own crimes – executions, for example, carried out on featureless terrain. If you could identify who and where, this could be evidence in a court of law. The shadow cast by a dead body was a strong indication of time of death.

At home, and surrounded by his daughter’s discarded toys, Higgins unearthed a number of scoops. He found weapons from Croatia in a video posted by a Syrian jihadist group. The weapons, it emerged, were from the Saudis. The New York Times picked up the story and put it on the front page – an indication of how armchair analysis could be as telling as dispatches from the ground.

Higgins documented the Syrian regime’s use of cluster bombs. He discovered that government soldiers were tossing DIY barrel bombs out of helicopters, and that rebels were fighting back around Aleppo with Chinese-made shoulder-launched missiles. His reputation spread. He launched a new investigative website: Bellingcat.

The idea was to consolidate pioneering online research techniques and to connect with a wider pool of international volunteers. In July 2014, three days after Bellingcat went live, a Malaysia Airlines passenger plane was blown out of the sky over Ukraine. Some 298 people – nearly 200 of them Dutch – died. The incident grew into Bellingcat’s first major investigation.

Higgins’s team discovered that the missile launcher had come from Russia’s 53rd anti-aircraft missile brigade, based in the city of Kursk. Video footage showed the launcher trundling across Russia as part of a military convoy. The system was filmed again by locals inside eastern Ukraine after MH17 was brought down, heading back to Russia with one of its missiles missing.

Bellingcat got bigger. One key figure was Christo Grozev, a fluent Russian-speaker and a Bulgarian from an anti-communist family. Grozev grew up in Plovdiv, Bulgaria’s second city; his father was fired from his job as a teacher for growing a hippie-style beard. Bellingcat joined forces with the Insider, an independent Russian news website run by Roman Dobrokhotov.

By summer 2018, the British police were confident they had identified the two Russian suspects who had tried to murder Sergei Skripal a few weeks earlier, in March. Skripal, a former officer with Russia’s GRU military spy agency, was poisoned in Salisbury, together with his daughter Yulia. The assassins’ names had not been made public. The hope was that they might travel to western countries where they could be arrested.

There were discussions inside the British government about what to do. One course was to demand their extradition – knowing Vladimir Putin would refuse, as he had with the killers of Alexander Litvinenko 11 years earlier. Another was to recognise that there was zero prospect of a criminal trial, and to publish concrete intelligence.

That September, prime minister Theresa May went with option two. She told the House of Commons that the two Russian assassins were Ruslan Boshirov and Alexander Petrov – adding that the police believed these names to be aliases. CCTV images from their trips to Salisbury were revealed. Also shown was the apparent murder weapon – a counterfeit perfume bottle containing the nerve agent novichok.

The new details were a boon for Bellingcat. During the next few weeks, its volunteers scurried all over the evidence. They would go on to inflict a series of humiliations on Russia’s GRU military intelligence spy agency that may have contributed to the fall of its chief, Igor Korobov.

In the 20th century, Soviet assassins were able to travel around Europe using fake passports. Their movements were seldom discovered. They may have been better, more professional spies – or lesser ones. It was an age before transparency.

The modern GRU was still using the old Soviet playbook when it came to covert operations such as the murder of enemies outside the country. These analogue plots now took place in a digital environment. GRU officers earned their spurs in the Soviet “near abroad” – in Tajikistan, Moldova or Ukraine, where there were few cameras to worry about, and not much of a CIA or other American presence.

Western Europe was different. Britain, in particular, was a counter-intelligence challenge. The UK had CCTV on every public corner – in railway stations, hotel lobbies and airports. Any passengers arriving on a flight from Moscow would be logged and filmed. A port-of-entry database was available to western security agencies.

Meanwhile, Russian markets sold CDs of mass official information: home addresses, car registrations, telephone directories and other bulk indexes. For £80 or so you could buy traffic police records. With the right contacts, and a modest cash payment, it was even possible to gain access to the national passport database.

Paradoxically, this low-level corruption made Russia one of the most open societies in the world. Corruption was the friend of investigative journalism, and the enemy of government–military secrets.

After the Metropolitan police published photos of Boshirov and Petrov, Bellingcat took up the hunt. It sought to unmask their real identities. The first step was to image search their photos via online search engines. This yielded nothing. They looked for telephone numbers associated with the two names. Nothing again.

And so the online investigators tried a deductive approach. They spoke to sources in Russia and asked where a GRU officer operating in western Europe was likely to have been trained. One answer was Siberia, and in particular the Far Eastern Military Command Academy in Khabarovsk, just across the Amur River from China. The men appeared to be in their late 30s or early 40s. This gave an approximate date of birth. . .

Continue reading. There’s more, and it’s interesting.

Written by LeisureGuy

24 June 2020 at 2:11 pm

Weeks after PTSD settlement, Facebook moderators ordered to spend more time viewing online child abuse

leave a comment »

Sam Biddle reports in The Intercept:

WITH THE INK still drying on their landmark $52 million settlement with Facebook over trauma they suffered working for the company, many outsourced content moderators are now being told that they must view some of the most horrific and disturbing content on the internet for an extra 48 minutes per day, The Intercept has learned.

Following an unprecedented 2018 lawsuit by ex-Facebook content moderator Selena Scola, who said her daily exposure to depictions of rape, murder, and other gruesome acts caused her to develop post-traumatic stress disorder, Facebook agreed in early May to a $52 million settlement, paid out with $1,000 individual minimums to current and former contractors employed by outsourcing firms like Accenture. Following news of the settlement, Facebook spokesperson Drew Pusateri issued a statement reading, “We are grateful to the people who do this important work to make Facebook a safe environment for everyone. We’re committed to providing them additional support through this settlement and in the future.”

Less than a month after this breakthrough, however, Accenture management informed moderation teams that it had renegotiated its contract with Facebook, affecting at least hundreds of North American content workers who would now have to increase their exposure to exactly the sort of extreme content at the heart of the settlement, according to internal company communications reviewed by The Intercept and interviews with multiple affected workers.

The new hours were announced at the tail end of May and beginning of June via emails sent by Accenture management to the firm’s content moderation teams, including those responsible for reviewing Child Exploitation Imagery, or CEI, generally graphic depictions of sexually abused children, and Inappropriate Interactions with Children, or IIC, typically conversations in which adults message minors in an attempt to “groom” them for later sexual abuse or exchange sexually explicit images. The Intercept reviewed multiple versions of this email, apparently based off a template created by Accenture. It refers to the new contract between the two companies as the “Golden SoW,” short for “Statement of Work,” and its wording strongly suggests that stipulations in the renewed contract led to 48-minute increases in the so-called “Safety flows” that handle Facebook posts containing depictions of child abuse.

“For the past year or so, our Safety flows (CEI,IIC) as well as GT have been asked to be productive for 5.5 hours of their day,” reads one email reviewed by The Intercept, referring to “Ground Truth,” a team of outsourced humans tasked with helping train Facebook’s moderation algorithms. “Over the last few weeks the golden sow, Accenture’s contractual agreement with Facebook, was signed. In the contract, it discussed production time and the standard that all agents will be held to.” Accenture moderators, the email continues, “will need to spend 6.3 hours of their day actively in production” — meaning an extra 48 minutes per day viewing the arguably most disturbing possible content found on the internet.

The email then notes that Accenture is “aligning to our global partners as well as our partners in MVW,” a likely reference to Mountain View, California, where, the email suggests, moderators were already viewing such content for 6.3 hours per day. It is understood, the email said, that there could be “one offs every now and then when you are unable to meet the daily expectation of 6.3″ hours of exposure, but warned against letting it become a pattern.

Pusateri, the Facebook spokesperson, told The Intercept, “We haven’t increased guidance for production hours with any of our partners,” but did not respond to questions about Accenture’s announcement itself. Accenture spokesperson Sean Conway said only that they had not been instructed to enact any change by Facebook, but would not elaborate or provide an explanation for the internal announcement.

Not only does the increase in child pornography exposure seemingly run afoul of Facebook’s public assurances that it will be “providing [moderators] additional support through this settlement and in the future,” it contradicts research into moderator trauma commissioned by the company itself. A 2015 report from Technology Coalition, an anti-online child exploitation consortium co-founded by Facebook and cited in Scola’s lawsuit, found that “limiting the amount of time employees are exposed to [child sexual abuse material] is key” if employee trauma is to be avoided. “Strong consideration should be given to making select elements of the program (such as counseling) mandatory for exposed employees,” the paper also noted. “This removes any stigma for employees who want to seek help and can increase employee awareness of the subtle, cumulative effects that regular exposure may produce.” The Accenture announcement, however, appears to fall well short of mandatory counseling: “Agents are free to seek out wellness coaches when needed,” the email states. A request for comment sent to Technology Coalition was not returned.

Accenture’s “wellness” program is a contentious issue for Facebook moderators, many of whom say such quasi-therapy is a shoddy stand-in for genuine psychological counseling, despite the best intentions of the “coaches” themselves. Last August,  . . .

Continue reading. There’s more.

FWIW, I make a small monthly contribution to The Intercept.

Written by LeisureGuy

18 June 2020 at 11:08 am

Murder exposes Facebook’s Boogaloo problem

leave a comment »

Judd Legum points out in Popular Information how Facebook promotes a violent right-wing anarchist group, a member of which murdered two law-enforcement offices. The column is definitely worth reading (as is Popular Information in general). A small snippet from a column that contains much more:

On June 2, shortly after Underwood’s murder, Facebook announced it would exclude Boogaloo pages from its recommendation engine. “We continue to remove content using boogaloo and related terms when accompanied by statements and images depicting armed violence. We are also preventing these Pages and groups from being recommended on Facebook,” a company spokesperson told Popular Information.

That doesn’t appear to be true. Boogaloo content continues to thrive on Facebook — and many Boogaloo-related pages continue to be included in Facebook’s recommendation engine.

Facebook still recommends the Boogaloo

Popular Information visited Boogaloo Side Quest, a popular Boogaloo page with more than 11,000 followers. Facebook then recommended two other Boogaloo pages: Boog Memes About Living The Dream (23,000 followers) and Rhett E. Boogie 2020 (13,700 followers).

Written by LeisureGuy

18 June 2020 at 9:46 am

Updated post

leave a comment »

I updated the Anki observations post with some illustrations.

Written by LeisureGuy

17 June 2020 at 2:19 pm

An Anki observation

leave a comment »

I continue to use Anki daily, and currently I’m using 13 decks, 3 of which I made and 10 downloaded from shared decks. These are all Esperanto decks, so there’s a fair amount of overlap, but each deck has some things the others lack. (I have deleted three decks simply because I had mastered all the vocabulary in them, and I’ve also go through all the cards in four of the current 13 decks so I get only review cards, no new vocabulary.)

You can see how many cards each deck will have for you, and nowadays the total number looks intimidating, but as it turns out many of these are cards I know so well that they should up only once every 2 to 3 months — it’s just that they’re spread out, so I’ll see a good number of them every day (and then not see them again for a few month).

The net result is that almost all the cards for a day are cards I know well, so I can go through them quickly and then not see them again for weeks. Of course, there are some stubborn cards I’m still working on, and occasionally a card I had indicated I knew will return and I’ll find that I don’t know it, so I click “Hard” and it comes back sooner.

I’m impressed by how well (and easily) it works. If there are any subjects that you must learn well and for which flashcards would be helpful, I highly recommend Anki — and do take a look at the shared decks. Those are not all good, but since you can readily edit any card, you can fix up minor errors and/or augment the information on the card.

UPDATE 17 June 2020: Here are the decks as of this morning:

I have no cards to review in Esperanto Affixes, and in Esperanto Correlatives, Recognize False Friends Like a Native!™, and Speak Esperanto Like a Native!™ 1 there are no new words to learn. I’ve been through those decks so it’s only review, and that goes quickly because I know all the words — the review is just reinforcement through spaced repetition.

The green numerals show how many I will review in each deck and the blue numeral shows how many new words are introduced, and I’ve left the default of 10 in place, though you can change that to match your ambition.

Here’s a typical recent word:

If I totally blank on the card, I click “Again,” otherwise I click on how easy it was. As cards are repeated the intervals get longer. Here are the choices for a brand new word:

And here are the choices for a word that I know well:

If I click “Easy,” I won’t see the word again for 4.2 months. Thus over time the daily review lessens dramatically. I average around 5 seconds a card.

Written by LeisureGuy

16 June 2020 at 7:18 pm

Excellent full-length documentary on AlphaGo and the match against the world champion

leave a comment »

I highly recommend this documentary even to those who do not play Go. I have no knowledge of (or interest in) football, but I loved the series “Friday Night Lights,” as so many do, not because of the football but because of the human drama. Football is really just the MacGuffin. The story is about the people, and it is absorbing because of that. So it is with this documentary.

Written by LeisureGuy

13 June 2020 at 9:24 pm

A Facebook account copied Trump’s words. Facebook censored the account for inciting violence.

leave a comment »

And yet Facebook will take no steps against powerful/wealthy people. Mark Zuckerberg is a despicable person — and does not embrace American values of equal treatment. David Gilbert reports in Vice:

UPDATE Thursday, June 11, 2020 6:56 p.m.: After the original story was published, the account holder confirmed that the post had been restored without warning. Facebook told VICE News the post had been deleted in error, but when asked how that error happened, Facebook failed to respond.

An account that copies word-for-word what the U.S. president posts on Facebook has been told to delete a controversial comment even though President Trump’s post was left untouched.

The SuspendThePres account was set up as an experiment to show how social media platforms treat high-profile public figures — specifically Trump — differently from ordinary users.

Last week a Twitter version of the account was temporarily suspended for posting the same message.

On Thursday it reported that Facebook had followed Twitter’s lead in forcing it to delete the controversial post, which includes the inflammatory phrase “when the looting starts, the shooting starts,” which has racist origins.

The account was warned that a repeat would trigger an automatic 24-hour suspension.

The reason Facebook gave the account holder, who has remained anonymous, was that the post “goes against Community standards on violence and incitement.” . . .

Continue reading. There’s more, and it’s worth reading.

Written by LeisureGuy

11 June 2020 at 8:56 pm

Maximizing benefits of Duolingo’s spaced repetition in language learning

leave a comment »

Anki explains well how two tactics maximize learning: active recall and spaced repetition. Quoting from that page:

Active recall

‘Active recall testing’ means being asked a question and trying to remember the answer. This is in contrast to ‘passive study’, where we read, watch or listen to something without pausing to consider if we know the answer. Research has shown that active recall testing is far more effective at building strong memories than passive study. There are two reasons for this:

  • The act of recalling something ‘strengthens’ the memory, increasing the chances we’ll be able to remember it again.
  • When we’re unable to answer a question, it tells us we need to return to the material to review or relearn it.

You have probably encountered active recall testing in your school years without even realizing it. When good teachers give you a series of questions to answer after reading an article, or make you take weekly progress-check tests, they are not doing it simply to see if you understood the material or not. By testing you, they are increasing the chances you will be able to remember the material in the future.

Spaced repetition

The ‘spacing effect’ was reported by a German psychologist in 1885. He observed that we tend to remember things more effectively if we spread reviews out over time, instead of studying multiple times in one session. Since the 1930s there have been a number of proposals for utilizing the spacing effect to improve learning, in what has come to be called ‘spaced repetition’.

Duolingo uses both active recall and spaced repetition

Duolingo structures its courses as a “tree” of skills, each skill shown as a disk with an icon. A skill has 5 levels, and after 5 levels the skill is completed (though you can do additional practice sessions if you want).

Each level comprises four to six lessons, typically six. Formerly, I would start a new skill and complete all five levels, then move to the next skill.

I finally realized that approach is bad because it undermines spaced repetition, which (along with active recall) truly solidifies learning. Active recall is built into every lesson of Duolingo, and Duolingo is also structured for spaced repetition. One obvious example of Duoling’s use of spaced repetition is how a mastered skill will occasionally, over time, be displayed as “broken,” to be fixed by completing a practice session.

The approach I had been using was counter to the idea of spaced repetition.

A better approach

The skills are displayed in rows on a language tree. When I finish a skill, I start a new available skill (a skill icon in color rather than grayed) by completing the first level in it. I keep 6-8 skills active, which amounts to skills in 3 or 4 rows (and perhaps not all skills in the rows are active because I haven’t started them).

I work sequentially through the skills I am currently working on, one level in each skill. That number seems to be about right: I return to the oldest open skill within a reasonable period of time to reinforce what I had learned earlier.

I work through the entire current set of 6-8 uncompleted skills (i.e., skills below level 5), completing one level in each skill before I repeat any skill. I don’t start a new skill until I complete one of the currently active skills.

The result is spaced repetition: I complete a skill level and move on to the next skill, returning later. Now that I’m doing it, I see that the levels seem to constructed with this approach in mind. My former approach amounted to cramming (as the night before a test), and that is not effective for long-term retention. Spaced repetition over time is.

I imagine most Duolinguists know this already, but I just figured it out and wanted to share it.

Update: Yep, this very approach was described in the Duolingo blog. Wish I had seen that post earlier. (Someone just sent me the link.)

Written by LeisureGuy

10 June 2020 at 12:14 pm

Calibre and your ebooks

leave a comment »

Calibre is a terrific program for managing your ebooks, and it gives you the ability to do some very nice things. Calibre is free, though donations are encouraged.

Lee Miller, in the Duolingo Esperanto Learner’s group on Facebook pointed out two good Esperanto resources available as PDFs: Teach Yourself Esperanto, John Cresswell, and Esperanto: Learning and Using the International Language, by David Richardson.

I downloaded the PDFs, and then wanted to put the Cresswell book on my Kindle. So I added the book (as a PDF) to my Calibre library. I then used Calibre to edit the metadata, correcting the author and title. They were both wrong, and since they are displayed in the library listing of the book, it was important that they be correct. Correcting them a snap: the metadata are displayed and you can easily edit the information.

I then used Calibre to covert the PDF to the format used by my eBook reader (AZW3 for my Kindle, but Calibre can also do MOBI, EPUB, and many other formats). I then copied the converted file to my Kindle.

Not quite Bob’s your uncle, but easy enough.

If you use ebooks, you should investigate Calibre. (The link is to a variety of YouTube explanations.)

Written by LeisureGuy

6 June 2020 at 10:08 am

Oral typos and autocorrect by the unconscious: Another Duolingo note

leave a comment »

One thing has become clear as I get my sea legs in Esperanto: the person listening — the auditor — contributes as much to oral exchanges as the speaker. Far from being a passive recipient of the speaker’s words, the auditor (once s/he is familiar with the language) unconsciously corrects the spoken equivalent of typos just as the reader does with written text — a process so unconscious that it becomes difficult to spot some typos, since the adaptive unconscious autocorrects such errors and thus renders them invisible to the conscious mind — see a famous example at the right. (And that is why copy editors get — or should get — the big bucks: they can still see what’s really there, not just what should be there.)

Just as we unconsciously correct and read right over typos because we know the language well enough to know what should be there, so also once we learn a spoken language we don’t notice so much the inaccurate word choices (“de” when “el” is meant) or mispronunciations (“kay” instead of “kai” (rhymes with “sigh”)). Our ears carry the sound, our brain corrects the meaning. We may wince at a mispronunciation (“toomeric” instead of “turmeric”), but we know immediately what is meant.

A person who is just starting to learn English, on reading the word “resaerch,” will try to figure out what it means, searching dictionaries for the word in a vain attempt to find the definition, while a person who knows the language — and the context of the sentence — will read “research,” perhaps not even noticing the error. (Writers, who know exactly what they meant to write, are notoriously poor at proofreading their own written work.)

I’ve observed as I listen to Esperanto dictation on Duolingo the same sort of thing. Initially I was hypersensitive to the slightest mispronunciation, which stymied my effort to understand the meaning. As the neural net of my brain has become more trained (through repetition and correction), those mispronunciations gradually fade from my attention, since I now “hear” (in my mind) the sounds that make sense. In effect, I adjust what I hear to match the most likely meaning even when it means ignoring some of the actual sounds that were made.

It’s similar to training an AI neural net to recognize cat pictures. At first the AI cannot tell the difference between a cat picture and a dog picture (or a cow picture, for that matter): a fur-covered four-legged animal with eyes, ears, a nose, and a tail: all much of a muchness. But with training — “No, not that one. Yes, that’s one. No. Yes. Yes. No. No. No. Yes. ….” — the AI soon is able to pick out cat pictures quite well.

Or an example from my own childhood: I still recall my mother in the grocery store asking me to get a head of lettuce (iceberg lettuce, all that we knew), and I came back with a head of cabbage. She laughed and showed me the difference, but I just couldn’t see it: they were both globes of green leaves, so how on earth could you tell them apart?

But experience — and neural net training — works and now I can pretty much distinguish cabbage from lettuce 9 out of 10 times, or even better.

What’s interesting is that my adaptive unconscious is learning Esperanto fast enough (daily practice of 2-3 hours now) that I can remember how some spoken sentences were unintelligible — and even had outright oral typos — now seem perfectly clear: because I now impose my own knowledge and understanding on what I hear and adjust to sounds to match what they should be.

Update: I overlooked the elephant in the room: the big contribution from the listener (as s/he learns the language) is the contribution of meaning, which is bigger and more important than the autocorrection of oral typos. There is no meaning in what hits the ear: that is just a bunch chopped up oral noises. For example, listen carefully to someone speaking, say, Basque and see what meaning you get from it. If you don’t know Basque, then you get from it no meaning at all. If you do know Basque, you get a lot of meaning, but the sounds alone do not contain the meaning — if the meaning were in the sounds, a Basque-ignorant person would get the meaning on hearing the sounds. A person who knows Basque and a person who does not both hear exactly the same sounds. One gets meaning, one does not. The difference is not in the sounds but in the listeners. The meaning is not in the sounds, but in the minds, that of the speaker and that of the listener.

Update again: And of the course the same is true of writing. The reader’s knowledge is a key factor in the writing being meaningful and not simply black marks.

이 구절은 명확한 의미를 지니지 만 독자가 표시의 중요성에 대한 지식을 제공하는 경우에만 가능합니다.

That has meaning, but not simply in itself. The reader must provide some of the meaning.

(I have a number of posts on what I’ve learned about using Duolingo effectively.)

Written by LeisureGuy

2 June 2020 at 10:14 pm

leave a comment »

Mantic59 of Sharpologist has launched a new site to serve as a communications vehicle from shaving vendors to potential customers: new products, special offers, discounts, and the like. Take a look.

Written by LeisureGuy

31 May 2020 at 10:07 am

Posted in Business, Shaving, Software

Training the adaptive unconscious

leave a comment »

I’ve mentioned before how learning a language (along with other kinds of learning) is very like how in AI one trains a neural network — because, of course, the brain is a neural network. And it seems clear that the part of the brain/mind being trained is the adaptive unconscious. Of course one must consciously learn the rules and the vocabulary and how to understand the spoken phrases one hears. But the idea is that it soon will not require conscious thought since the adaptive unconscious will take over the job, having through much repetition learned the patterns.

I was thinking this morning, as I found spoken sentences easier to understand and the choice of words to use in (say) telling time just came to me without my really having to think about it or even consciously understanding why I chose those particular words.It reminded me of when I began learning the Forth programming language. I had all sorts of mysterious crashes and malfunctions that gradually stopped happening without my ever understanding why most of them occurred. I think it was that through the experience of writing and debugging Forth I was training (through repetition) my adaptive unconscious and that as it absorbed the patterns/rules it simply directed me to choices that avoided the errors.

And part of that is teaching through mastery: you repeat an exercise in Duolingo until you get it right. If you get it wrong, the program shows you the correct answer and marks the error you made, but later in that session you will be presented the same exercise. If you get it right, great. If you get it wrong, you are again presented with it — repeatedly, until you get it right. That’s how you learn.

And it’s also how you work. When I made errors in programming, I had to keep at it until all the errors were fixed and the program ran properly. Repeating exercises until they are mastered is really the way to learn.

Update: Of course, performance skills — playing a musical instrument, for example, or acting a part in a play — are routinely taught with the practice of mastery: you work on the piece until you can do it right. In most math courses, on the other hand, your exercises are marked wrong and you do not get to repeat the exercise until you get it right (and thus learn).

Written by LeisureGuy

27 May 2020 at 1:22 pm

Language and machines, we and AI

leave a comment »

Written by LeisureGuy

25 May 2020 at 8:39 pm

Posted in Software, Video

‘Saluton!’: the surprise return of Esperanto

leave a comment »

Josh Salisbury writes in the Guardian in December 2017:

In the village of Barlaston, just outside Stoke-on-Trent, a strange language can be heard. It’s not the friendly Potteries dialect, but something that sounds a little like the lovechild of French and German. A Tudor house here is home to the Esperanto Association of Britain (EAB), which encourages people to learn the constructed language.

Esperanto summer schools were first established in Stoke in 1960, and were funded by the council for more than 50 years. The concept of an easy-to-learn, universal second tongue was energetically promoted, drawing prominent speakers to the area. And the city still bears traces of its Esperanto history. In Smallthorne, in the north-east of the city, drinkers can stop off on Esperanto Way to get a pint at the Green Star – the symbol of the Esperantist movement. A row of terraced houses a short walk away are located on Zamenhof Grove, named after the inventor of Esperanto, LL Zamenhof.

The language isn’t a relic; in fact, while exact figures are difficult to pin down, there has been a resurgence in people taking it up. The Esperanto Association has increased the number of beginners’ courses it offers four-fold to keep up with demand, says Viv O’Dunne, the charity’s operations and events director.

The “inner idea” of Esperanto, Zamenhof once said, was to promote world peace. A Jewish-Polish doctor born in 1859 in Białystok, now in Poland, Zamenhof grew up under Russian occupation. Violence between different groups was common – Białystok which was a melting pot of Protestant Germans, Catholic Poles, Orthodox Russians and Jews. While still a child, Zamenhof hit upon the idea that a constructed second language that was easy to learn and understand would allow people to talk as peers, rather than fight. In 1887, after tinkering away for more than 10 years, Zamenhof published his ideas in a pamphlet. By 1905, the fundamental rules of Esperanto had been established by a conference of speakers in France, and Esperantist groups began popping up across the world.

Tim Owen, education director for the EAB, gives me a crash course in what makes it straight-forward. “Probably the main factor is that you can acquire a huge vocabulary without knowing so many words,” he says.

All words ending -o are nouns, an -a ending is for an adjective, while -e denotes an adverb. He shows me that, for instance, “vidi” – meaning to see – can become “vido” for vision, “vida” for visual and “vide” for visually, concepts that require different words in English. If you need to find an opposite, you can add the “mal” prefix: “pura” is clean, “malpura” is dirty. These building blocks can help speakers learn new words very quickly. “It’s like working with a magic multiplier,” says Owen. The spellings are phonetic, there are no grammatical genders, verbs are strictly regular, and the vocabulary is a blend of European languages familiar to many. It’s child’s play to learn compared with my years of torturous high-school language lessons.

But isn’t a universal constructed language just a hobby for idealistic eccentrics? O’Dunne laughs, and concedes the stereotype might have a grain of truth: “We’ve still got those!” But, he says, there’s been a marked change in the demographic of attendees recently. “Over the last two or three years, there’s been much more interest from younger people who want to use it to travel and correspond … it just feels like it’s been rejuvenated,” she says.

At the charity’s headquarters at Esperanto House, there are rows of old academic tomes on the linguistics of constructed languages, but it’s the colourful translations of bestsellers and bright current affairs magazines that catch the eye. “[Esperanto is] a little bit geeky, but geeky is cool now, right?” says one recent convert.

Nineteen-year-old Sammy Kennedy, an aspiring photographer who works in retail in Manchester, is one of the young Esperantists helping the language shake off its niche image. He has attended events run by the EAB and has noticed more and more people taking an interest. Esperanto groups where he lives were defunct for ages, he tells me. “Now, there’s a new Manchester Esperanto group that meets up monthly,” he says.

Esperanto has become steadily more accessible largely thanks to the language-learning app Duolingo. The Esperanto course recently reached a million learners worldwide, more than are currently learning Hungarian or Czech on the site. There’s a dizzying array of other online options to help would-be learners, too. A few taps on the Amikumu (or “do the friendly thing”) app shows users local Esperantists to chat with, while numerous Facebook groups help beginners with vocab and grammar. Esperanto may have been the brainchild of a Polish doctor in the 19th century, but it has adapted for the 21st.

Simone Davis, a civil servant, began learning Esperanto online to distract her from painful chronic health conditions. She found that even at her most tired or ill, she could manage a lesson on her tablet. “One lesson easily becomes two or three and before I knew it I was hooked,” she says. In just over a year, she learned more Esperanto than she has French, despite taking French classes for five years.

It’s the values underpinning the language, as well as its ease, that drew in Davis. Esperanto is “a symbol of intentional goodwill towards others”, she says. Esperantists place a heavy value on the language being “neutral”, not belonging to one country. Many tell me they were inspired to pick up the books in response to what they see as rising isolationism in the era of Donald Trump and Brexit. Learning the language isn’t just a hobby, but a commitment to making connections across borders on a level playing field.

I initially write this off as simply a nice sentiment, but there’s plenty of practise behind the principle. O’Dunne shows me the Pasporta Servo, a pocket-sized directory of Esperantists all around the world. They offer fellow speakers a place to stay in their home country, often completely free of charge. For language enthusiasts under 25, the charity NoJef will pay for travel and accommodation for attendance at Esperanto-themed events.

The Pasporta Servo led 26-year-old James McMurray, a data engineer from Crawley, to make learning Esperanto his New Year’s resolution several years ago. He had first became familiar with the language while leafing through his grandfather’s books – he had become an Esperantist while stationed in India during the second world war. “I remember growing up and seeing his books in Esperanto, without being able to understand it, and his correspondence with people all over the world who may not speak English and be able to communicate,” McMurray says. He has since attended Esperantist music events in France, and met up with fellow learners in Prague and Malaga. His first serious relationship started through a shared interest in Esperanto. . .

Continue reading. There’s more.

On a personal note, after completing 12 lessons in the course, I have recently focused solely on the Duolingo course. As of this morning, I’ve done at least one lesson (and always in fact several lessons) a day for 36 days:

For what I learned about how to use Duolingo effectively, see this post.

This morning I also revised my two posts (this one and this one) on using Anki to reflect more discoveries on using it, and in particular to note the Duolingo shared decks and this quite valuable shared deck:

61,907 Esperanto dictionary entries in order of usage frequency in the Esperanto Wikipedia (dumped 2016-10-02). Dictionary entries come from Paul Denisowski’s Esperanto Dictionary Project (ESPDIC). To learn in order of frequency, you should set the options group for the deck in the “New Cards” section to have “Show new cards in order added” selected. Then the first few words to appear should be “la, de, en, kaj, esti”.

I submit that in this socially isolated, lockdown time, learning a language using Duolingo and Anki is a pleasant and useful way to pass some hours. And if you don’t know any foreign language, learning Esperanto first has been shown to greatly facilitate the learning of later languages.

I will say that after just over a month’s study I’m surprised by how much Esperanto I know. — or: Mi diros, ke post nur unu-monata studado mi estas suprizata per kiom da Esperanto mi scias. That I just wrote off the top of my head, using just what I’ve learned to date.


Written by LeisureGuy

24 May 2020 at 3:18 pm

Naomi Klein: How big tech plans to profit from the pandemic

leave a comment »

The Guardian republishes Naomi Klein’s article from the Intercept. It has a grim but probably accurate outlook on our future. Capitalism will always seek ways to profit from any event or trend or crisis. My saying this is not a reflex of anti-capitalism, any more than observing that a disturbed rattlesnake will strike or that rewards influence behavior. Capitalism is just a system to maximize profit, and that drives its behavior. Of course, if there is a calamity the capitalist response will be to seek a way to profit — that is simply what capitalism does.

That doesn’t mean that we simply lie down and let capitalism roll over us, but rather that we seek ways to control capitalize: exploit its strengths while blocking its destructive aspects. Government provides one means: laws and regulations that (for example) set severe limits on pollution — or (to take another example) that set minimum fleet mileage requirements. The regulations thus apply to all corporations in the sector, so no one corporation gets an unfair advantage, and then capitalism can evolve cost-effective solutions that meet the requirements.

Klein writes:

For a few fleeting moments during the New York governor Andrew Cuomo’s daily coronavirus briefing on Wednesday 6 May, the sombre grimace that has filled our screens for weeks was briefly replaced by something resembling a smile.

“We are ready, we’re all-in,” the governor gushed. “We are New Yorkers, so we’re aggressive about it, we’re ambitious about it … We realise that change is not only imminent, but it can actually be a friend if done the right way.”

The inspiration for these uncharacteristically good vibes was a video visit from the former Google CEO Eric Schmidt, who joined the governor’s briefing to announce that he will be heading up a panel to reimagine New York state’s post-Covid reality, with an emphasis on permanently integrating technology into every aspect of civic life.

“The first priorities of what we’re trying to do,” Schmidt said, “are focused on telehealth, remote learning, and broadband … We need to look for solutions that can be presented now, and accelerated, and use technology to make things better.” Lest there be any doubt that the former Google chair’s goals were purely benevolent, his video background featured a framed pair of golden angel wings.

Just one day earlier, Cuomo had announced a similar partnership with the Bill and Melinda Gates Foundation to develop “a smarter education system”. Calling Gates a “visionary”, Cuomo said the pandemic has created “a moment in history when we can actually incorporate and advance [Gates’s] ideas … all these buildings, all these physical classrooms – why, with all the technology you have?” he asked, apparently rhetorically.

It has taken some time to gel, but something resembling a coherent pandemic shock doctrine is beginning to emerge. Call it the Screen New Deal. Far more hi-tech than anything we have seen during previous disasters, the future that is being rushed into being as the bodies still pile up treats our past weeks of physical isolation not as a painful necessity to save lives, but as a living laboratory for a permanent – and highly profitable – no-touch future.

Anuja Sonalker, the CEO of Steer Tech, a Maryland-based company selling self-parking technology, recently summed up the new virus-personalised pitch. “There has been a distinct warming up to humanless, contactless technology,” she said. “Humans are biohazards, machines are not.”

It’s a future in which our homes are never again exclusively personal spaces, but are also, via high-speed digital connectivity, our schools, our doctor’s offices, our gyms, and, if determined by the state, our jails. Of course, for many of us, those same homes were already turning into our never-off workplaces and our primary entertainment venues before the pandemic, and surveillance incarceration “in the community” was already booming. But in the future that is hastily being constructed, all of these trends are poised for a warp-speed acceleration.

This is a future in which, for the privileged, almost everything is home delivered, either virtually via streaming and cloud technology, or physically via driverless vehicle or drone, then screen “shared” on a mediated platform. It’s a future that employs far fewer teachers, doctors and drivers. It accepts no cash or credit cards (under guise of virus control), and has skeletal mass transit and far less live art. It’s a future that claims to be run on “artificial intelligence”, but is actually held together by tens of millions of anonymous workers tucked away in warehouses, data centres, content-moderation mills, electronic sweatshops, lithium mines, industrial farms, meat-processing plants and prisons, where they are left unprotected from disease and hyper-exploitation. It’s a future in which our every move, our every word, our every relationship is trackable, traceable and data-mineable by unprecedented collaborations between government and tech giants.

If all of this sounds familiar, it’s because, pre-Covid, this precise app-driven, gig-fuelled future was being sold to us in the name of friction-free convenience and personalisation. But many of us had concerns. About the security, quality and inequity of telehealth and online classrooms. About driverless cars mowing down pedestrians and drones smashing packages (and people). About location tracking and cash-free commerce obliterating our privacy and entrenching racial and gender discrimination. About unscrupulous social media platforms poisoning our information ecology and our kids’ mental health. About “smart cities” filled with sensors supplanting local government. About the good jobs these technologies wiped out. About the bad jobs they mass produced.

And most of all, we had concerns about the democracy-threatening wealth and power accumulated by a handful of tech companies that are masters of abdication – eschewing all responsibility for the wreckage left behind in the fields they now dominate, whether media, retail or transportation.

That was the ancient past, also known as February. Today, a great many of those well-founded concerns are being swept away by a tidal wave of panic, and this warmed-over dystopia is going through a rush-job rebranding. Now, against a harrowing backdrop of mass death, it is being sold to us on the dubious promise that these technologies are the only possible way to pandemic-proof our lives, the indispensable keys to keeping ourselves and our loved ones safe.

Thanks to Cuomo and his various billionaire partnerships (including one with Michael Bloomberg for testing and tracing), New York state is being positioned as the gleaming showroom for this grim future – but the ambitions reach far beyond the borders of any one state or country.

And at the dead centre of it all is Eric Schmidt.

Well before Americans understood the threat of Covid-19, Schmidt had been on an aggressive lobbying and public-relations campaign, pushing precisely the Black Mirror vision of society that Cuomo has just empowered him to build. At the heart of this vision is seamless integration of government with a handful of Silicon Valley giants – with public schools, hospitals, doctor’s offices, police and military all outsourcing (at a high cost) many of their core functions to private tech companies.

It’s a vision Schmidt has been advancing in his roles as chair of the Defense Innovation Board, which advises the US Department of Defense on increased use of artificial intelligence in the military, and as chair of the powerful National Security Commission on Artificial Intelligence, or NSCAI, which advises Congress on “advances in artificial intelligence, related machine learning developments and associated technologies”, with the goal of addressing “the national and economic security needs of the United States, including economic risk”. Both boards are crowded with powerful Silicon Valley CEOs and top executives from companies including Oracle, Amazon, Microsoft, Facebook and of course, Schmidt’s former colleagues at Google.

As chair, Schmidt – who still holds more than $5.3bn in shares of Alphabet (Google’s parent company), as well as large investments in other tech firms – has essentially been running a Washington-based shakedown on behalf of Silicon Valley. The main purpose of the two boards is to call for exponential increases in government spending on research into artificial intelligence and on tech-enabling infrastructure such as 5G – investments that would directly benefit the companies in which Schmidt and other members of these boards have extensive holdings.

First in closed-door presentations to lawmakers, and later in public-facing opinion articles and interviews, the thrust of Schmidt’s argument has been that since the Chinese government is willing to spend limitless public money building the infrastructure of high-tech surveillance, while allowing Chinese tech companies such as Alibaba, Baidu and Huawei to pocket the profits from commercial applications, the US’s dominant position in the global economy is on the precipice of collapsing.

The Electronic Privacy Information Center (Epic) recently got access, through a freedom of information (FOI) request, to a presentation made by Schmidt’s NSCAI in May 2019. Its slides make a series of alarmist claims about how China’s relatively lax regulatory infrastructure and its bottomless appetite for surveillance are causing it to pull ahead of the US in a number of fields, including “AI for medical diagnosis”, autonomous vehicles, digital infrastructure, “smart cities”, ride-sharing and cashless commerce.

The reasons given for China’s competitive edge are myriad, ranging from . . .

Continue reading. There’s much more.

Written by LeisureGuy

17 May 2020 at 9:51 am

The usefulness of Calibre

leave a comment »

I was browsing through a list of adventure stories, and found this entry:

E. Nesbit’s children’s fantasy adventure The Phoenix and the Carpet (1904). Nesbit’s writing is extraordinary, because: she is drily witty, in a way that makes her writing as entertaining to adults as to children; she was politically progressive, so there’s not quite as much to cringe at as we find in her contemporaries’ writing; and her conception of how magic and the everyday world might interact — the brilliant idea that magic has strict rules which need to be puzzled out, painfully — would prove influential on everyone from P.L. Travers to Edward Eager, C.S. Lewis, Diana Wynne Jones, and J.K. Rowling. Here, siblings Cyril, Anthea, Robert, and Jane, whom we first met in Five Children and It (1902), discover a flying carpet and an ancient Phoenix — a magnificent, vain creature which expects to be worshipped by modern Londoners. Over the course of several adventures, the children wear out the carpet’s fabric — which causes it to malfunction, entertainingly. There’s a burglar, a buried treasure, and a church jumble sale that goes entertainingly awry. Will the children figure out the best possible use for the carpet before it’s too late? Fun facts: The final installment in the Psammead series is The Story of the Amulet (1906). I’m also a fan of Nesbit’s Bastable series, her House of Arden series (including 1908’s The House of Arden), and her other children’s novels — including The Railway Children (1906) and The Enchanted Castle (1907).

E. Nesbit’s name rang a bell, as did The Enchanted Castle, which I’m sure I read some decades ago. A quick search took me to a site with the full text, but I wanted to read it on my Kindle, not my computer. I saw that you can download the text in MOBI format, but that download is chapter by chapter, not the entire book. The PDF is the entire book, but that’s not a good format for the Kindle. The EPUB file is the full novel, but Kindle doesn’t do EPUB (nor, for that matter, does it any longer do MOBI — it insists on AZW3).

However, the (free) program Calibre, an ebook management program, can convert files from one format to another, and in particular from EPUB to AZW3. And before you could say “Jack Robinson,” I had the file downloaded, converted, and transferred to my Kindle. (I admit you would have to speak v-e-r-y s-l-o-w-l-y to still be saying the name when I had finished, but it did take very little time.)

An aside on the etymology of the phrase “before you can say ‘Jack Robinson'”?

Uncertain. There is some speculation that this is a reference to Sir John Robinson, a Lieutenant at the Tower of London around 1600, but there is nothing known about him that is associated with speed (Samuel Pepys called him as “a talking bragging bufflehead.”), and the phrase does not appear in print until 1778.

If you read ebooks, Calibre is worth a look.

Written by LeisureGuy

11 May 2020 at 10:31 am

Posted in Books, Software, Technology

Two types of epidemiology: Models v. Evidence

leave a comment »

Jonathan Fuller writes in the Boston Review:

The lasting icon of the COVID-19 pandemic will likely be the graphic associated with “flattening the curve.” The image is now familiar: a skewed bell curve measuring coronavirus cases that towers above a horizontal line—the health system’s capacity—only to be flattened by an invisible force representing “non-pharmaceutical interventions” such as school closures, social distancing, and full-on lockdowns.

How do the coronavirus models generating these hypothetical curves square with the evidence? What roles do models and evidence play in a pandemic? Answering these questions requires reconciling two competing philosophies in the science of COVID-19.

In one camp are infectious disease epidemiologists, who work very closely with institutions of public health. They have used a multitude of models to create virtual worlds in which sim viruses wash over sim populations—sometimes unabated, sometimes held back by a virtual dam of social interventions. This deluge of simulated outcomes played a significant role in leading government actors to shut borders as well as doors to schools and businesses. But the hypothetical curves are smooth, while real-world data are rough. Some detractors have questioned whether we have good evidence for the assumptions the models rely on, and even the necessity of the dramatic steps taken to curb the pandemic. Among this camp are several clinical epidemiologists, who typically provide guidance for clinical practice—regarding, for example, the effectiveness of medical interventions—rather than public health.

The latter camp has won significant media attention in recent weeks. Bill Gates—whose foundation funds the research behind the most visible outbreak model in the United States, developed by the Institute for Health Metrics and Evaluation (IHME) at the University of Washington—worries that COVID-19 might be a “once-in-a-century pandemic.” A notable detractor from this view is Stanford’s John Ioannidis, a clinical epidemiologist, meta-researcher, and reliable skeptic who has openly wondered whether the coronavirus pandemic might rather be a “once-in-a-century evidence fiasco.” He argues that better data are needed to justify the drastic measures undertaken to contain the pandemic in the United States and elsewhere.

Ioannidis claims, in particular, that our data about the pandemic are unreliable, leading to exaggerated estimates of risk. He also points to a systematic review published in 2011 of the evidence regarding physical interventions that aim to reduce the spread of respiratory viruses, worrying that the available evidence is nonrandomized and prone to bias. (A systematic review specific to COVID-19 has now been published; it concurs that the quality of evidence is “low” to “very low” but nonetheless supports the use of quarantine and other public health measures.) According to Ioannidis, the current steps we are taking are “non-evidence-based.”

This talk of “biased evidence” and “evidence-based interventions” is characteristic of the evidence-based medicine (EBM) community, a close relative of clinical epidemiology. In a series of blog posts, for example, Tom Jefferson and Carl Heneghan of the Oxford Centre for Evidence-Based Medicine similarly lament the poor-quality data and evidence guiding action in the pandemic and even suggest that lockdown is the wrong call.

In the other corner, Harvard’s Marc Lipsitch, an infectious disease epidemiologist, agrees that we lack good data in many respects. Countering Ioannidis’s hesitation, however, Lipsitch responds: “We know enough to act; indeed, there is an imperative to act strongly and swiftly.” According to this argument, we could not afford to wait for better data when the consequences of delaying action are disastrous, and did have reason enough to act decisively.

Public health epidemiologists and clinical epidemiologists have overlapping methods and expertise; they all seek to improve health by studying populations. Yet to some extent, public health epidemiology and clinical epidemiology are distinct traditions in health care, competing philosophies of scientific knowledge. Public health epidemiology, including infectious disease epidemiology, tends to embrace theory and diversity of data; it is methodologically liberal and pragmatic. Clinical epidemiology, by contrast, tends to champion evidence and quality of data; it is comparatively more methodologically conservative and skeptical. (There is currently a movement in public health epidemiology that is in some ways closer to the clinical epidemiology philosophy, but I won’t discuss it here.)

To be clear, these comparisons are fair only writ large; they describe disciplinary orthodoxy as a whole rather than the work of any given epidemiologist. Still, it is possible to discern two distinct philosophies in epidemiology, and both have something to offer in the coronavirus crisis over models and evidence. A deeper understanding of modeling and evidence is the key not only to reconciling these divergent scientific mindsets but also to resolving the crisis.

section separator


Public health epidemiology uses theory, especially theory from other health sciences like microbiology, to model infection and understand patterns and causes of disease. Many of the epidemic models that the public and public health researchers alike have been voraciously consuming—including models produced by Imperial College London that informed the U.K. and U.S. coronavirus response—are SIR-type models. The theory underlying these models is old, originating in the Kermack–McKendrick theory in the 1920s and ’30s, and even earlier in the germ theory in the second half of the nineteenth century. The SIR framework partitions a population into at least three groups: those who are susceptible to future infection (S), those who are currently infectious (I), and those who have been removed from the infectious group through recovery or death (R). An SIR model uses a system of differential equations to model the dynamics of the outbreak, the movement of individuals among the various groups over time.

Other models in the SIR family add additional groups to these three basic ones, such as a group for those who are infected with the virus but not yet infectious to others. Agent-based models also represent infection dynamics (how the number of cases changes over time), but they do so by modeling behaviors for each member of the simulated population individually. Curve-fitting models like the one used by the IHME are less theoretical; they extrapolate from previous infection curves to make predictions about the future. All these different models have been used in the COVID-19 pandemic. The diversity of approaches, along with divergent estimates for model parameters, partly explains the range of predictions we have seen.

Public health epidemiology also relies on  . . .

Continue reading. There’s much more.

Written by LeisureGuy

8 May 2020 at 2:18 pm

Understanding Duolingo better — and treating it as a game

leave a comment »

Here’s a screenshot of where I am in my Duolingo Esperanto course:

The image shows a tower labeled “2.” That is the second checkpoint. At a checkpoint you take a test to determine whether are you ready to advance beyond the tower to the next collection of skills. I was, and I’m now working through the new part of the tree. If you are using Duolingo to brush up on a language you know, you can click one of the checkpoints to take the test and move on from there — for example, you might click checkpoint 3 and see if you pass. Obviously, if you’re learning a language new to you, you just start at the beginning and move through the tree, level by level.

Each disk in the tree is a “skill.” Each skill has 5 levels, and each level is reached by going through a series of (brief) lessons. I have completed the first two skills (Family 2 and Useful), the 5 in the crown showing that all 5 levels have been achieved. When level 5 is reached, the skill becomes golden with a gold halo. (After some time has elapsed, the skill will show a jagged break, but you can return to the skill and and with a practice session repair the break — spaced repetition.)

I have completed only the first level (the “1” in the crown) of Home 2, Action 2, and Numbers 2. (The “2” is because those skills are follow-ups to earlier skills  Home 1, Action 1, and Numbers 1.)

I have not completed any of Dates, Occupations, and Requests, but they are “unlocked,” so I can start lessons in them.

Affixes 1, Describe, and Directions still are locked, and those won’t be unlocked until I complete more or the available skills.

What I have been doing was to work my way straight through all the levels in a particular skill before moving to the next. Now, however, I stick with one skill until I complete a level (or two), but then I move to another skill to complete a level there. I limit myself to working on 6-8 skills at a time. By moving among those 6-8 skills it is more interesting and also improves learning and recall when I return to an uncompleted skill and do another level (until it finally is complete).

Although you can leave a skill before a level is complete, stopping after any lesson, I prefer to stick with the skill until I’ve completed a level.

Update: This approach is exactly what Duolingo recommends. /update

Progress is measured by XP: “experience points.” Completing a lesson means you get 10 XP, with a bonus of up to 5 XP depending on how few errors you made. (If you do provide incorrect answers, you are shown the correct answer and then later you are asked the question again (and again) until you get it right, the idea being mastery. (Anki does the same sort of thing.)

In “Settings,” you can set your daily target (and you get awards for having an unbroken streak of days in which you completed at least one lesson). Here is my setting:

Screen Shot 2020-05-06 at 8.29.26 PM

I chose “Intense,” but it’s really not all that intense. For the first couple of weeks, I was doing around 500XP per day, but now I aim for 200-300XP — which amounts to 2-3 levels.

It’s worth noting that Anki’s shared decks for Esperanto include a couple of decks for Duolingo vocabulary. I use those to lock down my vocabulary knowledge. See this post for more information on Anki.

Now that I better understand how it works, I find it more interesting. And I have to say that it’s a great indoor activity for quarantine.

In fact, Greg Hullander has an interesting blog post on treating Duolingo as a game (in which he also describes the skill tree).

More of my Duolingo discoveries in this post.

Written by LeisureGuy

6 May 2020 at 7:25 pm

%d bloggers like this: