31/01/2017

notnot notable words


  • pshrink (n.): an odd version of shrink, psychiatrist, with a silent p. Possibly a shibboleth used by anti-psychiatry people. Used in both Tom Clancy and Asimov!

  • gynotician (n.): gyno-politician. .

  • IOW (prepositional modifier.): in other words.

  • mantissa (n.): the fractional part of a floating point number. see also significand.

  • ablative analysis (n.): machine learning model evaluation by gradual deconstruction. "Remove components from your system one at a time, to see how it breaks."

  • meet cute (n.): an unusual and romantic first encounter (between the protagonist of a romcom).

  • online judge (n.): an online programming competition. In competitive job markets like India they seem to be gaining serious economic weight. The practice is applicable to code in particular, since it is effortlessly remote, objectively scrutable...

  • Spad (n.): a Special Advisor (to a minister of the UK government ).

  • Growing programs from data (v.): twee name for machine learning in Domingos (2012), an incredible paper, with maybe three important insights per A4 page.

  • lemmatise: to turn a word into its morpheme. (The wiki page does not mention the word "morpheme", and so must expend 150 words doing what my preceding sentence did.)

  • Abelianisation (verbal n.): the mapping between a non-Abelian group and its smallest normal subgroup. (Which subgroup is always Abelian.)

  • digraph (n.): 1) a pair of letters. 2) a directed graph, an abstract collection of pairs with one-way roads between them.

  • open core: open source software, but with a superior paid version offered by the company that maintains it. For 'Freemium' is a grave insult in this corner of the world. MySQL is the big one, but the Databricks distro of Spark.

  • recurse: to use recursion on. Amusing pun for anyone who has written recursive code: cursed again.

  • induce: to work out by induction. Compare deduce.

  • headfake: obvious once you know:



Also:


Done in January 2017


  • Awesome New Year's day with dears. Pride: 1
  • Saw all my family and other family-like peeps. Pride: 2
  • Made my first public predictions. Pride: 3.
  • Made a fancy LaTeX CV. Pride: 3
  • Started a fancy new self-hosted blog. Pride: 3
  • Got into a cheap but excellent diet: power smoothies and Peter Specials. Pride: 1
  • Got a promotion. Pride: 4
  • Wrote about political controversy. Pride: 3
  • Wrote about cultural anthropology (again). Pride: 2
  • Wrote about the noun "intellectual". Pride: 2
  • Wrote a link archiving script. Pride: 2
  • Started going to pub quizzes with my darlings. Pride: 1
  • Read Chaos at last. Pride: 1
  • Read Age of Em. Pride: 2
  • Maths assignment #4: physical modelling, harmonic motion, normal modes and eigenvectors. Pride: 2
  • Presented DBSCAN and backpropagation to C-levels. Pride: 2
  • Helped my mate look for jobs. Pride: 1
  • Planned a lot, recorded a lot, e.g. this. Pride: 2
  • Rewrote my quotations extensively. Pride: 2.
  • Long-distance relationship with Pesukaru continues very well. Pride: 5.


22/01/2017

is Yitang Zhang an intellectual?


Yitang Zhang in conversation


An intellectual is a person who engages in critical thinking... about the reality of society, and proposes solutions for the normative problems of society, and thus gains authority as a public intellectual. Coming from the world of culture... the intellectual participates in politics, either to defend a concrete proposition or to denounce an injustice, usually by producing or by extending an ideology...
- wiki

Yitang Zhang achieved great progress on one of the outstanding problems in number theory, the twin primes conjecture. But he doesn't hype his work, hasn't set himself up as a celebrity, and I imagine it is impossible to get a political statement out of him.

Remarkably, our primary usage of the noun "intellectual" wouldn't include Zhang. It is something like:

  1. intellectual: A brainy outspoken celeb. A famous person who broadcasts their opinions on serious social matters. (Particularly if they're involved in the public speaking circuit, particularly if they speculate about politics.)

Scholars (especially scientists or mathematicians) who don't debate in public and don't mouth off about, say, elections or creationists on Twitter are excluded from this sense. (As if said mouthing off was a more creditable act of intellect than proving the deep connectedness of the Monster group and modular functions.)

This is pure Two Cultures bullshit: a mental land grab by arts people. (They pulled off a similar annexation of the idea of creativity, around the late C19th.) All scientists are knowledge workers, and that's the sense of "intellectual" I'd set as primary, if language was in my power.

It is no deficiency of Zhang that 1) the discourse can't follow number theory and 2) that he is a comically and beautifully diffident man.

Let's see when "intellectual" came in as a proper noun:

(hover over the blue line to see the year)


Doesn't tell us much. We can get some evidence for the arts colonisation of the term as follows:


Which is a pleasingly strong signal: it is born in the 30s, takes hold in the 40s and increases through the mid-C20th, the long years of 'Theory' in the nonsciences.


Some wider senses for the term:
  1. intellectual: Any scholar. Anyone whose labour is primarily knowledge-based.
  2. intellectual: Anyone whose interests include researching and forming reasoned positions on classically academic things.
  3. intellectual: Anyone whose interests include researching and forming reasoned positions.

So usage (1) covers a really tiny fraction of intellectually active people, and a fraction even of people who spend all their time forming and sharing opinions about the world. This is maybe because "intellectual" (1) is a contraction of "public intellectual" - which is maybe derived from the Leninist concept of a vanguard intelligentsia: basically just propagandists, whose role was to steer the educated public. (So the hypothesis is: social commentary was the raison d'etre of the Marxist intelligentsia, and so their descendents and the descendents of their opponents have inherited this distorted and limited idea of intellectuality.)

I'm not insisting that usage (4) is best just because it's biggest. Consider the little essays at the start of mainline TVTropes pages (e.g. these brief histories of animation): they definitely are research, definitely the work of critical minds. Are they intellectuals? Wiki editors often display enormous erudition. "Hobbyist" or "fan" can cover some of this group, if we must leave "intellectual" with its present high status. But expansion is necessary too.

(Dissemination does seems important: a full-blown intellectual should publish, and engage with incompatible views. So it's only the hostile takeover by social theorists and celebrity culture I'm actually opposing.)


21/01/2017

doing a bit


"Yeah, the wedding was last week."
"Oh! Congratulations."
"Job's just begun."



"If the backlash against fake news succeeds, will the fakers aim for credibility - say mixing in two truths for each lie?"
"We have those already, they're called newspapers."



"Your identity has been stolen."
"Amazing! You mean there's another one of me running around?"
"No: stolen. You are not you."



"People find clarity repugnant. Look at maths."
"People find obscurity repugnant. Look at modern art."



"I'm going up to read philosophy."
"Oh; my first degree was philosophy."
"Where did you go?"
"Aberdeen."
"Why did you pick
there?"
"The honest answer is, Because they gave me money to, and I felt guilty about leaving my mother on her own. The answer you want to hear is, Have you heard of the Northern Institute it briefly did cutting-edge work in maths and language I also like Luca Moretti Steve Bruce Bill Naphy.

At any rate, your question reflects an unworthy fixation on status, when the only distinctive value of university philosophy is that its uncommercialisability lets you transcend gross social pressures in favour of substance: nobody cares enough to exploit or degrade it. Try 'So where do you stand on [Question]?' if you ever want to not be a prig."



"I was trying to think of myself as a slacker surrounded by mediocrities."
"But a slacker's a
de facto mediocrity."
"So I saw."



"So, after all, James will be the software architect for this project."
"How senior is he?"
"Do you mean, how good is he?"
"No, I mean: 'am I being insulted with non-prestigious staff?'"



"Our guide is, simply, ordinary experience —"
"— suitably formalized, of course."
"Of course."


11/01/2017

notable ew it's in my ear


  • radiomics (n.): machine learning for diagnosis x-rays automagically.

  • strace (n. and v.): the venerable low-level Unix debugger.

  • thot (US n.): Initialisation of 'that ho over there'. op. cit. A synonym for 'slut' but with class sneering tacked on.

  • decision stump (n.): a decision tree with only one rule.

  • heisenbug (dev n.): software bug which disappears when you call someone else over to check it out

  • bivalvegan (n. or adj.): Vegan who eats nonsentient animals. Coined by Diana Fleischman, who has some arguments in favour that I actually buy. I don't eat bivalves because I think you can get the gains from unpretentiousness from freeganism, and without gambling with the pain of others.

  • in pectore (Latin adj.): 'in one's breast'; secretly. Term for the pope's secret appointment of cardinals. See also Italian, in petto, English cards held close to his chest.

  • the vol (Finance n.): volatilty

  • mandatory conference (n.): a private legal summons: "talk to me or else"

  • high-speed idiocy (n.): Gleick's term for digital computing

  • determinate physical mechanism (n.): Newell's gloss on "machine". I don't have a better definition, this just struck me as jarring, too broad.

  • digital (proper n.): As in


    So disgusting.

03/01/2017

Predictions for 2017

Politics
  1. Total terrorism death toll < 30,000. 70%.
  2. Total malarial death toll > 350,000. 90%.
  3. Total malarial death toll < 500,000. 80%.
  4. No terrorist attack causes >100 deaths in the developed world (HDI > 0.8): 80%
  5. No terrorist attack causes >500 deaths in the less developed world: 70%.
  6. No terrorist attack outside of Iraq and Syria causes >300 deaths: 55%.

  7. No country will leave the Eurozone. 80%
  8. Fewer refugees admitted to European countries in 2017 than 2016. 80%
  9. Front National will gain 30+ seats in the French legislative election. 80%
  10. Le Pen will lose the presidential election. 55%
  11. Front National will not end up in power. 80%
  12. Partij voor de Vrijheid will gain 5+ seats in the Dutch election. 70%
  13. PVV will not end up in power. 80%.
  14. Alternative für Deutschland will gain 40+ seats in the German election. 70%
  15. AfD will not end up in power 70%.
  16. No second Scottish independence referendum announced. 80%
  17. There will be no formal agreement on UK membership of the common market, post-Brexit. 80%
  18. The UK triggers Article 50. 80%
  19. No snap UK general election. 55%
  20. Conditional on no GE, no new leadership challenge to Corbyn. 70%
  21. May's approval rating above 30%. 55%

  22. China will not enter recession. 80%
  23. No large revolt against North Korea’s government (riots or civil war). 95%
  24. No major revolt (Tiananmen Square+) against Chinese Communist Party: 95%
  25. Putin’s domestic approval will remain above 50% in whatever reputable polls exist there. 90%
  26. Pro-Russian forces will continue to hold Donetsk and Luhansk. 80%.
  27. Ethiopia will see GDP growth of >= 7%. 70%
  28. Joseph Kabila will not be in power. 55%
  29. Merkel will win the Nobel Peace Prize, with Abdullah II of Jordan and others. 55%
  30. Erdogan will not be so honoured. 70%
  31. The Iran nuclear deal will hold (i.e. no Iranian uranium enrichment) 70%
  32. Raqqa will remain under the control of ISIL. 55%.
  33. Syrian ceasefire will not last. 70%
  34. Assad will remain President of Syria. 70%.
  35. Total media mentions of ISIL continues to decline. 70%

TRUMP
  1. Trump will not be impeached. 90%
  2. No new prosecution of Hillary Clinton. 90%
  3. US will not get involved in any new war with death toll of >100 US soldiers: 70%
  4. Trump's first state visit will be to the UK. 55%
  5. Trump will end the year above 35% approval. 70%
  6. Trump’s approval will be below 60% throughout the year. 90%.
  7. Mexican wall will become a fence, but symbolic structures will be under construction. 70%
  8. TPP will not be ratified this year. 80%
  9. US will not withdraw from TPP. 55%
  10. Deportations from the US will not increase by more than a third. 55%
  11. Obamacare will not be substantially cut (i.e. no, >10% spending cut or coverage retraction). 70%
  12. F-35 program will not be curtailed. 70%

Other
  1. 2016 will remain the hottest year on record. 90%
  2. Bitcoin will end the year above $400. 70%
  3. Brent crude will finish above $60 per barrel. 55%
  4. EMDrive thrust will not be positively replicated by an independent party. 80%
  5. There will be a larger DDoS attack than the October Dyn attack. 70%
  6. No attempt of a human head transplant. 80%
  7. No human head transplant results in a conscious subject. 95%
  8. No verified quantum supremacy this year. 55%
  9. The SAV001 HIV vaccine will show significant results, Cohen's d > 0.3. 55%
  10. At least one PlayStation-level data breach. 80%
  11. "Data science" hype will continue to increase, as measured by Google Trends hits / world job openings. 70%
  12. A CRISPR clinical trial will be conducted on humans in the US. 80%
  13. Conditional on it occurring, it will have disappointing results. (Cohen's d < 0.3)70%

Personal
  1. I will be in the same job. 80%
  2. I will not be offered a Summer Fellowship at CEA. 70%
  3. I donate 20% of pre-tax income to a new basket of charities. 90%
  4. I will reduce my reading to <50 books. 90%
  5. I will post fewer posts. 80%
  6. I will complete two decent-sized software projects. 55%
  7. I will join a choir. 55%
  8. I will not play Civ. 70%
  9. I will use Airbnb at least once in 2017: 60%

  10. I will get 70% of these predictions right. (Correct sign) 70%
  11. I will get a Brier score of > 0.2. 55%
  12. I will be 5%+ underconfident on these predictions. 55%





Known other predictors: Anders Sandberg, Scott Alexander, John Vert, Stephen Carter,