Let's watch: Black Mirror, Hated in the nation

Black Mirror is a good hard sci-fi series with significantly less appeals to emotions than average sci-fi. We'll watch season 3, episode 6, Hated in the Nation (2016), by Charlie Brooker et al.

Also it's as long as a usual movie (90 min).

Cheatsheet

• Karin Parke: old detective, DCI, darker shorter hair than Blue. Traditional, thinks Blue is naive, not good at modern tech.
• Blue Colson: young detective, TDC, lighter longer hair than Karin, sometimes wears glasses. Digital forensics expert.
• TDC: Trainee Detective Constable
• DCI: Detective Chief Inspector
• 
  
• Jo Powers: first death, writer.
• Tusk: second death, rapper.
• Clara Meades: third death, took bad selfie at war monument.
• 
  
• Colony Collapse Disorder: strange phenomenon where a lot of worker bees in a colony disappear, causing the colony to fail.
• ADI: Autonomous Drone Insect, robot bees, made in the Project Swarm.
• Granular: company that makes the Project Swarm ADIs.
• Rasmus Sjoberg: engineer in Granular, head of Project Swarm.
• Vanessa Dahl: woman with Maud hairstyle in Granular.
• Tess Wallender: exemployee of Granular, was cyberbullied.
• Garett Scholes: flatmate of Tess Wallender, likes Tess. Wrote The Teeth of Consequence
• 
  
• NCA: National Crime Agency, police department of the UK
• GCHQ: Government Communications Headquarters, intelligence department of the UK
• Shaun Li: NCA agent, fat man in suit.
• POI: person of interest

Bayesian detective

There were two detectives. It took me a bit to figure out what the matter is with the two detectives. One was old and the other was young. First death happened during the afternoon and Jo died.

The Old insisted that Jo was killed by Jo's husband, and the Young thought it was some of the haters online that killed her. This is a classic Bayes scenario with two hypotheses:

1. "Jo was killed by Jo's husband"
2. "Jo was killed by some of the online haters"

Given either hypothesis, the outcome would be Jo's death, which is one of the main evidences. There are other evidences such as the online hates, Jo's husband also slashed in the stomach, etc. But I'm a novice in Bayesian reasoning so I won't incorporate these.

The evidence "Jo died" unfortunately has no bearing on the hypotheses likelihoods, so they are dismissed. 

To use Bayesian probability to calculate how likely the hypothesis is, we need prior. But how? We use a very crude estimation of Solomonoff prior (quantitative Occam's Razor).

1. "ADI kills randomly"
2. "ADI kills only people hated online"

Note that I tried to use the most succinct language possible, because in Solomonoff prior, only the most succinct description counts.

The second hypothesis uses 3 more words, so, by Claude Shannon's estimation, one English word contains about 10 bits, which gives about $2^{-10} = 1e-3$, thus the prior odds ratio is
$$\text{hypotheses }1 :\text{hypotheses }2 = 1 : 1e-9$$
Then, after the deaths of Jo and Tusk, two people hated online, the odds are suddenly changed.

If ADI kills randomly, then since there are about 50 million people in England, the probability of Jo, then Tusk, being killed, is $(5e7)^{-2} = 4e-16$.

If ADI only kills people hated online, then we need to calculate how many people could be hated online in England at any one time. Now, this is a matter of Fermi estimate again, but I'm not sure about that. I'll just guess the number is 100? It seems impossible, even for the whole of English internet, to hate intensely over 100 people at the same time.

So that gives this table:

	random ||	only hated online
prior	1	1e-9
Jo and Tusk killed	4e-16	1e-4
posterior	4e-16	1e-13
	1	250

So, the odds of hypothesis "only hated online" is in favor by 250, which is as we expected: after the two deaths, the hypothesis "only people hated online are killed" is far more likely than "random people are killed".

#DeathTo Japan from China

It reminds me a bit about something in China, see this and that. In brief, nationalistic Chinese people like to say slogans that amount to #DeathTo all Japs.

Since Mao, words like xiāomiè 消灭 ("eliminate; extirpate") have crept further into daily-life Chinese than they had ever been before, and in that sense they are not literal.  But the very "normalization" of bloodthirsty language probably makes violence more possible, too. Sapir-Whorf had a point, I think.

PRC-era rhetoric has a well-established reputation for dehumanization and licensing all manner of mass savagery... That said, this anti-Japanese "genocide" rhetoric seems new, with no real historical analogue, and it strikes me as more or less the way Chinese "nationalism" expresses its desire to avenge what Imperial Japan once did to China.

And don't get me started on how nationalistic Chinese people like to call Americans "American imperialists". They say it's just a joke, but is it? Or is it nationalistic bullying?

Snarky remarks here

Bees buzz, what the buzz feed.

A few possible ways to deal with the ADI:

• EMP bomb.
• Live in a submarine or a bunker or abroad.
• Wear a face mask.

"Not a performance" might be the biggest hypocritical remark in the entire movie lmao

The theory of Garett, as detailed in The Teeth of Consequence, is that, by baiting out people who likes to group-cyberbully others, then letting ADI to kill them, it would remove cyberbulling traits from the gene pool. But more seriously, he probably wasn't that dumb. So probably it was just a social performance rather than a genetic cleansing, which makes the "not a performance" wrong.