Somehow both dry and heavy on intuition, folk ML. Stuff which you'll actually use. I've brushed up against most of it before (: I've called most of itSomehow both dry and heavy on intuition, folk ML. Stuff which you'll actually use. I've brushed up against most of it before (: I've called most of it from the safe distance of a nice Python library before), but it took a second pass and doing all the exercises to click.
To actually learn (grok) something, you need
1. To do it, not just read about it 2. To read it several times 3. To feel challenged but not overwhelmed by it
And 2&3 conflict.
(Most books don't have a natural do-operator. How do you do a novel? I make do with these reviews; others do fanfiction and probably get the same benefit.)
Kind of annoying that the figures are never next to their discussion. And I was hoping this would make me like R but I can't and I don't. But good....more
Was looking for a qualitative introduction to convey something of the excitement and philosophical importance - the art of discovering anything which Was looking for a qualitative introduction to convey something of the excitement and philosophical importance - the art of discovering anything which isn't bleedin' obvious! nor knowable apriori! This has bits of that ("Statistics is applied philosophy of science"; "it is the technology for handling uncertainty") but is still too dry to recommend as a first exposure.
He diagnoses the worst parts of university teaching: hand calculations, canned inference, and the (exhausting, interminable) bag-of-tools approach, rather than computers and The Framework. But the latter have steep learning curves. I think the biggest thing missing is simple tailoring of datasets: let them pick something they care about to study, to learn how to study on.
Lots of ML methods covered, without a single mention of the phrase "machine learning". This is fair enough if you consider how much of (enterprise) ML hype is just rebranded 40 year old stats.
Hand notes the origins of the field - as State-istics, i.e. as the beginning of bureaucracy and surveillance. But he doesn't feel the tension of this fact: that it helped to transform us, for good and ill, into legible people.
One dodgy idea: he claims that numbers offer a more direct apprehension of reality than words, that they're realer. But this isn't why they're better: they're better because they're more sensitive - it's at least possible for them to track any size change in the world, while words are mostly stuck to medium-sized dry goods - and because they are easier to spot errors in....more
Short, friendly, smooth, repetitive. First ML book where I didn't feel dumb.Short, friendly, smooth, repetitive. First ML book where I didn't feel dumb....more
Exercise 3.11... Is the lawyer right to imply that the history of wife-beating does not point to Mr S’s being the murderer? Or is the lawyer a slim
Exercise 3.11... Is the lawyer right to imply that the history of wife-beating does not point to Mr S’s being the murderer? Or is the lawyer a slimy trickster? If the latter, what is wrong with his argument?
[Having received an indignant letter from a lawyer about the preceding paragraph, I’d like to add an extra inference exercise at this point: Does my suggestion that Mr. S.’s lawyer may have been a slimy trickster imply that I believe all lawyers are slimy tricksters? [Answer: No]].
Actually readable, actually tractable. We could not afford to lose him.
The exercises are graded 1-5 by difficulty. If it helps, notice that difficulty 3 is the kind of thing which stumped a young David Mackay for "some time" (a week?) (cf. Exercise 3.3).
Casual, brief and witty theoretical economics. Actually a unification of economics, political science, and this other thing.
The prose is deceptively Casual, brief and witty theoretical economics. Actually a unification of economics, political science, and this other thing.
The prose is deceptively plain, like say Anscombe. There's basically no data, it's all intuition pumps. (But I share most of his intuitions, which surprised me.)
Some key premises: * Nonmarket forces matter. * Dissent helps us. It is extremely rich data for anyone who cares about doing good work, and so lasting. * If there were only rational (critical and disloyal) customers, the economy would collapse. * Exit (market pressure) and Voice (nonmarket pressure). * (TODO (...) )
I've never seen such a clear analysis of nonmarket forces - and since I find them ugly and hopeless traps for activists, it's good to be challenged.
He cracks open a massive neglected topic: we had studied macroeconomic slack (depressions, where your whole system is suddenly sluggish and unresponsive) extensively, but not microeconomic slack - Why don't firms maximise output, in healthy macro regimes? Because they're made of individuals who don't? Because of internal conflict requiring suboptimal compromises? Because it serves a latent function?
There's also a surprising focus on social / personal decay - the idea that our new productive power allows for an unprecedented range of decadence, failures, mutation load... There's certainly some evidence for that. But also selection-as-repair and post-traumatic growth:
Firms and other organisations are conceived to be permanently and randomly subjected to decline and decay. This radical pessimism, which views decay as an ever-present force constantly on the attack, generates its own cure; for as long as decay is hardly in undisputed command at all times, it is likely that the very process of decline activates certain counter-forces.
Read it slowly. I found it really clear, but I don't know if that's because I'm already fairly grounded in economics.
He has a touching pride in political science, in a 1950s Apollonian way, and much of this criticises economists' costly bias in favour of exit:
I hope to demonstrate to political scientists the usefulness of economic concepts and to economists the usefulness of political concepts. This reciprocity has been lacking in recent interdisciplinary work as economists have claimed that the concepts developed for the purpose of analyzing phenomena of scarcity and resource allocation can be successfully used for explaining political phenomena as diverse as power, democracy and nationalism. They have thus succeeded in occupying large portions of the neighbouring discipline.
Has the field since done much to vindicate this pride?
The best argument against Exit, which I usually favour, is that it speeds up and deepens a group's decay: the people who get sick of drama and then leave are likely to be the more thoughtful and virtuous.
Worth serious study, but if all you take from this review is taking both market and nonmarket forces seriously, and thinking about when to use each, then it's still likely to be some help....more
Fun with a serious scientific mission. The expected titillating facts are present (how many people have tried anal? How many people are gay? What's tyFun with a serious scientific mission. The expected titillating facts are present (how many people have tried anal? How many people are gay? What's typical?) but there's also an intro to the many difficulties of social science and a history of sexology in here. You learn why you should admire (but not trust) Ellis, Hirschfeld, Kinsey, Masters & Johnson, Hite...