My review of CS183 - Stanford - Blake Masters' Notes:
How should one assess this book which is not even a book but some dude's ridiculously thorough noMy review of CS183 - Stanford - Blake Masters' Notes:
How should one assess this book which is not even a book but some dude's ridiculously thorough notes from a seminar that prominent venture capitalist Peter Thiel taught at Stanford a few years ago (Crown is turning it into a real book, but the free version is here)?
Consider a 2 x 2 matrix. (The 2 x 2 matrix appears to be a pet rhetorical exercise of Mr. Thiel.)
[image]
On the vertical axis you have a most basic assessment of the veracity of the content. Everyone has forgotten about James Frey but they remember that most non-fiction is some percentage fiction. "Truth" gets lost in service to narrative, or through just plain incompetence or irrationality.
On the horizontal axis we have engagement—although the cynical cliche is that there is an inverse relationship between entertainment value and honest journalism or scholarship, we see that sometimes that is not true.
I would put this book squarely in the middle of all four quadrants.
[image]
It is an awkward Frankenstein monster of oftentimes ugly but undeniable truths, advice no less eccentric for how clearly hardwon it was, some really kookoo ideas, inquisitiveness regarding those who find such ideas kookoo, some fascinating reminiscing and reverse engineering of the early PayPal days, a celebration of the virtues of taking a rigorous, skeptical, yet contrarian VC approach to any risky venture, including life itself, and original and spot-on digressions on topics like advertising, why cleantech has failed, and the changing zeitgeist of luck vs. determinism.
It's worth reading, but probably best with a buddy or book club so that you can talk about the many weird parts and unique insights together....more
When I bought this, even though it had been published in 2002 I thought that it had been updated regularly. It hasn't, and although useful there is noWhen I bought this, even though it had been published in 2002 I thought that it had been updated regularly. It hasn't, and although useful there is no reason why the ebook should be so expensive. I will have to do a lot more reading on the internet to get up to date....more
"I don't know. It's just a trick, you know? It's like when I was... I wrote this story on Paris, and I'd never been there. You don't have to— It's jus"I don't know. It's just a trick, you know? It's like when I was... I wrote this story on Paris, and I'd never been there. You don't have to— It's just a trick." —Rain to Gabe, Husbands and Wives, written and directed by Woody Allen
Gabe, a writing professor played by Allen, has just peppered Rain, his precociously talented student played by Juliette Lewis, with some extremely flattering inquiries into her writing methods and background. Rain's embarrassed humble-brag of a response only serves to further her mystique. She's not the creepily ambitious kind of precocious, she's the effortlessly good kind.
And that's what we wonder about our computers, isn't it? What gallons of dystopian, man vs. the machines ink have been spilled over? Are the smartComputerPhoneMachines doing these remarkable things to our economy and cultural and political apparatuses because they are just that good, or, embedded in all the assumptions of what makes good software and a good tech company are there some pretty creepy ambitions too?
I got through this book in a week, and I think I understand most of it. (It's just a trick, you know?) The book did have its difficult moments. Because it uses very few practical examples, it was difficult to extrapolate the significance of the variations between some of the data structures, but many of the concepts were very elegant, even beautiful, to me once I was able to extrapolate the significance on my own. I took copious notes, and there were a few times when I wrote "nifty" or "cool" in the margins of my notes.
There's one problem called the "Minimum Spanning Tree" which aims to find the shortest existing network of connections between a group of points. Here's the script I wrote in JavaScript based on Kruskal's Algorithm (a little clunky because if JavaScript can do sets I'm not yet aware of how) to find the Minimum Spanning Tree for a given set of eight islands with multiple ways to connect:
var vertices = 8; var C = [];
function falseSet(number) { var array = []; for (count = 0; count < number; count++) { array[count] = false; } return array; }
for (n = 1; n <= vertices; n++) { C[n] = falseSet(vertices); C[n][n-1] = true; }
var Q = [ Set of several bridge lengths between 8 vertices in format [length of bridge,[id number of vertex A,id number of vertex B]] ]; Q.sort();
var T = [];
for (m = 0; m < Q.length; m++) { var v = Q[m][1][0]; var u = Q[m][1][1]; if (C[v] !== C[u]) { T.push(Q[m]); var thisCluster = falseSet(vertices); for (i = 0; i < vertices; i++) { if (C[v][i] || C[u][i]) { thisCluster[i] = true; } } for (i = 0; i < vertices; i++) { if (thisCluster[i]) { C[i+1] = thisCluster; } } } }
I first tried to figure out the problem by hand hoping that would lead me to the right JavaScript methodology faster than translating the book's pseudocode solution. I drew the first four islands and their potential bridges in my notes, but my meatspace CPU failed to distinguish a usable pattern even at that small scale.
So, I turned to my HTML editor and went back and forth between Kruskal's and another solution Bruvka's Algorithm (which had at first seemed the easier to implement of the two) before finally translating Kruskal's with some success. And now I can apply Kruskal's to my hand-drawn four-island graph easily in my head.
Because now I know the trick.
So this stuff is all just logic (a great thing), probability (a very good thing), and efficiency (a powerful thing). It's breaking problems down into subproblems and piecing them back together so that over the long-term and at great scale the problems can be solved with magnitudes greater efficiency. This is such a key concept in computer science that it even has a name, "Dynamic Programming."
How can any of this be wrong, let alone bad for us or creepy?
Life is a problem: it's made up of mostly suffering and then we die. Certain subproblems of life can be optimized methodically, and tricks are welcome in such cases: the articulation and complication of dark truths about human imagination; instantaneous graphical, text, and quantitative data on potential new commutes.
But even the post-modernists knew that tricks alone were not enough: they also gave us an appreciation for process over product. Engineers know this, at least those who truly understand Lean Startup do. Minimum Viable Product is as French New Wave of a philosophy as you get. (Is "Minimum Viable Product" the least beautiful rewrite of "the journey is the destination" possible, or what?)
And process is the definition of inefficient. If a process were truly efficient it wouldn't exist. You would have an idea and then there would be a product in O(1) time. No engineer would release something created on that schedule. So why hurry everyone else along?
Christopher Isherwood wrote A Single Man at a rate of eight pages a month, and not because he used a typewriter (I know that this is not provable, but I just know it's true). He wrote at a rate of approximately 950 characters per day, or less than seven Tweets!
Of course, I read it at a much faster rate than that, and it was sent from Amazon's servers to my Paperwhite at a much, much, much faster rate than that, thanks to several of the algorithms and data structures in this book and the hard-working engineers who figured out how to implement them for this very purpose for millions of people. Around the world. Simultaneously. (Cool! Nifty!)
However, clearly the actual writing of the book had an expensive O(something very large) runtime. It was, in fact, supported by a teaching position at a state-funded institution and the post-war housing boom in California (also subsidized by the American government).
As many know, bipolar disorder is a chronic psychological illness characterized by alternating episodes of mania and depression. During manic episodes, bipolars experience an increase in energy and a decreased need for sleep, with some forgoing sleep altogether. Attention span is low, and judgment impaired. Behavior may become aggressive, intolerant, or intrusive. He or she may feel out of control or unstoppable, or as if he or she has been "chosen" and is "on a special mission," or have other grandiose or delusional ideas.
I'm not trying to make any person or persons seem crazy so much as trying to show that some really extreme efficiencies, such as those created by manic episodes, don't come free. There is no such thing, I believe, as chronic mania. Absent of pharmaceutical stimulation, mania occurs only in conjunction with alternating episodes of major depression.
One thing that I was surprised to learn from this book is how imperfect some of the math is. There are some algorithms and data structures that depend on patterns of probability, even randomization, to increase efficiency in a real world with unpredictable inputs. That is, their worst-case run-times are actually quite high, but their average runtimes over time fall within desirable ranges. They work like an insurance system: the super fast searches, insertions, and deletions make up for the unexpectedly slow ones over time. Goodrich and Tamassia call this "amortization," but it sounds an awful lot like redistribution of resources to me.
Algorithms are Socialist Democrats!
Some of them, anyway.
"I wrote this story on Paris, and I'd never been there. You don't have to— It's just a trick."
Americans love freaks like Rain. We love natural talent, effortless success. But, paradoxically, we also don't trust it. We ask it what its parents do for a living, where it grew up (Gabe asks Rain both of these things), we try to reduce the improbable input to if-then logic. But this bipolar love-hate relationship with outliers does not make for a robust system. If redistribution is good enough for our algorithms, it's good enough for us....more
"We do know that Siren Servers can die. It happened to Lehmann Brothers... Individual Siren Servers can die and yet the Siren Server pattern persevere"We do know that Siren Servers can die. It happened to Lehmann Brothers... Individual Siren Servers can die and yet the Siren Server pattern perseveres, and it is that pattern that is the real problem. The systematic decoupling of risk from reward in the rising information economy is the problem, not any particular server."
I'm sure much savvier readers and technologists than me will roll their eyes at a neologism like "Siren Servers," Jaron Lanier's nickname for the entities (Amazon, Facebook, Google*, government surveillance apparatuses, etc. along with some less-obvious manifestations like Goldman Sachs, Wikipedia, Wal-Mart, and health insurance companies) that use powerful computing and the perfect bet that is Moore's Law to consolidate power and/or wealth that they haven't really earned under their own control, and, most dangerously, radiate risk outward.
"Not long after Jaron’s birth, his parents abandoned their last name, Zepel, for the less Semitic-sounding Lanier, after Sidney Lanier, a nineteenth-century poet and flutist, whom Ellery admired."
Jaron Zepel's parents knew the power of names in the game of self-preservation in a cruel world. I would argue that the entities listed above are also familiar with the power of names.
Names are data.
Not that I have to explain this to this audience, but here are some qualities of the Sirens of Homer: -part human female -part bird -talented musicians who perform for free -immortal -people starved to death rather than leave them
This book is as dark and filled with ancient, Jungian forms of dread as its central metaphor implies. It is filled with practical suggestions that I think have been too casually dismissed by other reviews. His suggestion of paying people for when their data is useful doesn't pass the creepiness test that he himself discusses at length, but Zepel is making a larger argument that these companies will allow business and technical forms of accountability as long as they are applied across the market to all of the big players. That will take a lot of political will, but less than I think we fear.
As emotionally taxing as it is, there were long swaths of this book that I loved reading the way I love reading a good, ambitious shaggy novel. Zepel writes passionately about putting the needs of technology below the needs of people, using specific examples from the sci-fi canon to illustrate what he means in an unemotional manner.
The biggest flaw of the book is that it does not engage the research that Clay Shirky discussed in Cognitive Surplus, that suggests that paying people to do things that they do to feel accomplished or connected changes how they feel about these things, for the worse.
I've read one Shirky, one Zepel, and another book called Database Nation, and it is striking how much personal experience colored these apolitical, non-fiction books. Shirky once underestimated the appeal of self-designed web page hub Geocities, and promised never to make that mistake again. Zepel attended one too many fundraisers for music producers and musicians needing medical coverage and finally decided that a Siren Servers-run economy, at least in current, hegemonic incarnation, is, like Communism, an experiment responsible for too much bad to be good. They are writing what they know. Attempting to be as machine-like and objective as possible, but at the same time their humanity, appreciation for irony, and respective wealth of real life experience show through.
They are biased, but that is a strength not a weakness.
Bias means you have skin in the game.
Zepel is arguing, essentially, that we should start coding what we know, not just what is easiest and most popular, and appreciating engineers and tech organizations that demonstrate an appreciation for human scale, irony, and having skin in the game.
This is a tough review to structure and write. First, I am not the ideal reader for this book. Although I am learning web design and care about accounThis is a tough review to structure and write. First, I am not the ideal reader for this book. Although I am learning web design and care about accountability and embrace technology in my professional life, I am not trying to launch anything at scale and certainly not with VC money.
But the single most striking thing about Eric Ries's book, original recipe Lean Startup (Ries was the Editor of Lean Analytics, but the publisher here is O'Reilly not Crown Business), was how readily applicable his concepts were to so many professional ventures besides VC-backed startups.
I didn't see that same hitting zeitgeist on the funny bone here.
So many of the chapters, under the guise of offering specifics and a rigorous approach to startup management, do not offer specifics so much as undisciplined definitions of business jargon, unmemorable introductions to market research practices that have been around for decades (cohort analysis ftw!), anecdotes and case studies that rarely come to life (some are not even data or Lean related), and some very poorly designed and written charts.
I do not think that the authors closed the sale on the One Metric That Matters. It's not clear that they even made a concerted effort to try. There is utility in their startup stages framework, but whereas Build, Measure, and Learn are three clear, active verbs in a similar vein, Empathy, Stickiness, Virality, Revenue, and Scale don't have much in common besides all being nouns evocative of the stages they represent. Empathy is a quality of the founder, while Stickiness (someone please, please invent a new term for whatever this is) and Virality are qualities of the product, and Revenue and Scale are qualities of the business. Do you see where my frustrations are stemming from? There is intellectual clarity and rigor lacking when even the five core concepts are not consistent with each other.
I actually think The Signal and the Noise is a better book to read to spur creative thinking about milking the data beast in the age of the internet.
You can read both if you want, but this book ain't cheap, and here are some examples of the Susan Miller-esque insights to be found within—Susan Miller-esque because if you want it to all make sense enough it will:
"It's better to make big bets, swing for the fences, try more radical experiments, and build more disruptive things, particularly since you have fewer user expectations to contend with than you will later on."
"A fundamental element of any pricing strategy is elasticity: when you charge more, you sell less; when you charge less, you sell more."
"The reality is you'll quickly adjust the line in the sand to your particular market or product. That's fine. Just remember that you shouldn't move the line to your ability; rather, you need to move your ability to the line."
-
PREVIOUSLY:
I love Lean Startup, but this is how it really goes sometimes, amirite?
Very appropriate to read this on my evil data-collecting devices (although I bought this title direct from O'Reilly and then accidentally deleted the Very appropriate to read this on my evil data-collecting devices (although I bought this title direct from O'Reilly and then accidentally deleted the cloud copy 50% of the way through, so I'm not sure that my highlights got stored) and then, upon finishing, enter still more data into yet another database (THIS ONE—I sometimes forget this is a database, and now Mr. Garfinkel's made-up story about an online friend who turned out to be a robot makes me question who among my Goodreads network are robots. I would say Gunn, because what human would admit to reading The Divine Secrets of the Ya-Ya Sisterhood?, but I know that he is just a strange human and not a robot.)
This book is a treasure trove of sci-fi story plots, even 10 years after its original publication, but Garfinkel is clearly a left-brain guy. (That is code for: this book could have been better written.) It is quite interesting, but not the awesomely definitive account I was expecting from the striking title and subtitle....more
Loading...
