Pepe Escobar is both a polymath and highly insightful. This is a must read, about Cathy (mathbabe) O'Neil on mathematical modeling and computer algorithms as fashioning "weapons of math destruction." Here are some highlights, but the whole thing is worth a read. It's pretty short and engaging.
… a book such as Weapons of Math Destruction, by Cathy O’Neil (Crown Publishing) becomes as essential as the air that we breathe.
O’Neil is the real deal; PhD in Math in Harvard, former professor at Barnard College, former quant at a hedge fund before reconverting as a data scientist, and a blogger at mathbabe.org.
Mathematical models are the engines of our digital economy. That propels O’Neil to formulate her two critical insights – which may startle legions who regard machines as simply ‘neutral’.
1) “Math-powered applications powering the data economy [are] based on choices made by fallible human beings”.
2) “These mathematical models [are] opaque, their workings invisible to all but the highest priests in their domain: mathematicians and computer scientists. Their verdicts, even when wrong or harmful, [are] beyond dispute or appeal. And they tend to punish the poor and the oppressed in our society, while making the rich richer”.
Thus, O’Neil’s concept of Weapons of Math Destruction (WMDs); or how destructive math models are accelerating a social earthquake.
O’Neil extensively details how destructive math models now micromanage vast swathes of the real economy, from advertising to the prison system, not to mention the finance industry (as in all the after effects of the never-ending 2008 crisis).
These math models are essentially opaque; unaccountable; and target above all “optimization” of the (consuming) masses.
A golden rule is – what else – to follow the money. As O’Neil puts it,
for “the people running the WMDs”, their “feedback is money”; “the systems are engineered to gobble up more data and fine-tune their analytics so more money will pour in».
Victims – as in Obama administration drone strikes – are mere “collateral damage”.…Two maxims of math modeling, which sum up most of political economy:
- Cui bono? "To whose benefit?" with respect to modeling assumptions.
- GIGO with respect to data.
It's not just about money but also power:
Most Americans – not to mention most of Facebook’s 1.7 billion users around the world – ignore that Facebook tinkers with the news feed; people actually believe that the system instantly shares anything that is posted with their community of friends.
Which brings us, once again, to the key question in the news front. By tweaking its algorithm to model the news people see, Facebook now has all it takes to game the whole political system. As O’Neil notes, “Facebook, Google, Apple, Microsoft, Amazon have vast information on much of humanity – and the means to steer us in any way they choose”.
Their algorithms, of course, are strategically priceless; ultimate, non-transparent, trade secrets; “They carry out their business in the dark”.Counterpunch
Playing Algorithm ’n Blues
Pepe Escobar
13 comments:
Theyre applying stochastic mathematics to what are deterministic subjects....
I dont think this is a problem with the "math" per se....
Looks like that's another book I'm going to have to read. I agree 100% with what Escobar says she thinks.
Matt, stochastic mathematics says what was true in the past will be true (easily predictable) in the future, like astronomical calculations. They can tell within a second when an eclipse of the moon will happen in 2300 AD just as they can recreate the astronomical record of eclipses of the moon going back several millennia. They're axiomatic.
I think what O'Neil is saying that the future cannot be predicted on the probability distributions of the past, but that these scribblers of algorithms are behaving as if they could; furthermore, they are injecting their 'desired future' by circumscribing choices and restricting people's ability to wider choices.
And btw, the first time I heard the expression "Weapons of Math Destruction" was out of Paul Davidson's mouth in his undergraduate lunch lecture at the Univ. of Chicago, three or four years ago. (on YouTube)
Amazes me how many people still use facebook and find it relevant. Maybe it's like AOL and people will cling to it for years as it slowly declines.
Facebook is no technological internet invention. It is merely the digital version of AT&T's circa-1910 party line. Everything controlled from the hub. Big whoop.
The Nosy Parker who controlled who could call whom in a rural area was the all-powerful gossip who could shape what the community was thinking, if she desired. Party lines still existed in the 60s and 70s, fercrissake. My aunt had one on her farm, and she found it anti-deluvian after her urban life.
I get sneered at for not using Facebook, as if I'm sort of digital troglodyte. I tell them I was a Twitter user when it wasn't called Twitter--started with the letter "O", April 2006--and when I stopped using Twitter it only had 150,000 members! I think I'm going to go back one day
I have a visceral dislike of Zuckerburg. He stole the idea and the tech from those VossWhatever twins, then got Larry Summer to back his theft. Eben Moglen, a man I consider a genius and ethical, says Zuckerburg "richly deserves bankruptcy." Well Peter Thiel made sure that wouldn't happen.
MRW astronomical calculations are not stochastic... they are a perfect example of a deterministic calculation...
We dont say "well... we dont know where the sun is going to be but we do know that it will be dark for 50% of a day...."... "we dont know when the sun is going to set but it will set once per day..."
We know the angular velocities and distances, etc and we do deterministic calculations...
Here:
"In mathematics and physics, a deterministic system is a system in which no randomness is involved in the development of future states of the system."
https://en.wikipedia.org/wiki/Deterministic_system
Youre doing the same thing as Reihart/Rogoff....
"well, 90% of nations with debt:gdp ratios over 1 go hyper-inflation!"....
Its the same old story where 'correlation is not causation..."
To make a prediction, you need to have developed a functional equation and then use derivative action of that function in order to make predictive statements... using ANY ex post data is trial and error... its not even an "educated guess"...
youre saying that if you flip a coin and get 5 heads in a row, then the chances are that the next flip is going to be a heads too...
I joined Facebook just so I could join a Hi Fi an online Hi-Fi club. I scrambled all my information and said I was a professor of mathematics and was 96 years old. I succumbed and put photos of myself on it, though, and some family pictures so my brother could see them, and then some old dodgy school friends caught up with me who I wasn't interested in meeting.
Apart from food I buy almost everything from Amazon and eBay, so the amount of information they have on me is frightening. I even bought a kindle which now has 100's of books on it all on leftist causes, and 100's more which are samples and so the range is huge.
I was looking at an article on scam artists the other day, which I might post that here, but apparently all the information we put online is up for grabs.
People apparently will post photos online of what they have in their houses and then tell their friends that they are on holiday for a few weeks.
Then everything we write here is logged somewhere, and all the ' Likes ' on YouTube, and all the debates were might enter.
Chris Hedges says he not radical enough to be of any real interest to the authorities so I guess that applies to us here as well.
Apparently Google has CIA connections. I don't know if that is a conspiracy theory, or not, but I wouldn't be surprised if was set up for mass surveillance.
"I think what O'Neil is saying that the future cannot be predicted on the probability distributions of the past"
The future is millions of artificial neural nodes talking to each other. They construct their own probability weights dynamically - and have the advantage that the decision traces can be checked for correspondence with reality.
They work on internalised rules and incomplete data sets. The emergent properties from their interaction will be much more enlightening than linear algebra.
Apparently Google has CIA connections.
Apparently.
How the CIA made Google
Inside the secret network behind mass surveillance, endless war, and Skynet—Part 1
BTW, that is a 4 part series
Here is a link to all four parts.
MRW astronomical calculations are not stochastic...
You're right. I was wrong.
I keep confusing the two. Astronomical calculations are ergodic, not stochastic. That's what makes them axiomatic. My apologies.
Post a Comment