»You can sign up to receive each day’s Start Up post by email (arriving at about 0800GMT each weekday). You’ll need to click a confirmation link, so no spam.«
A selection of 10 links for you. Use them wisely. I’m @charlesarthur on Twitter. Observations and links welcome.
Every time there’s a mass shooting or terror event, due to the subsequent backlash, this YouTube conspiracy genre grows in size and economic value. The search and recommendation algorithms will naturally ensure these videos are connected and thus have more reach.
In other words, due to the increasing depth of the content offerings and ongoing optimization of Youtube’s algorithms, it’s getting harder to counter these types of campaigns with real, factual information.
I hate to take the dystopian route, but YouTube’s role in spreading this “crisis actor” content and providing these thousands of false videos is akin to a parasite-host relationship. This genre of videos is especially troublesome, since the content has targeted (individual) effects as well as the potential to trigger mass public reactions.
The view count for 50 of the top mass shooting-related conspiracy videos is around 50 million. Not every single video overlaps directly with conspiracy-related subjects, but it’s worth pointing out that these 8842 videos have registered almost four billion (3,956,454,363) views.
Contrary to my earlier remarks on Twitter about YouTube’s algorithm getting “gamed,” I’m no longer sure. The only gaming here appears to be using tragic events for automated content monetization.
Perverse incentives. Read on for more.
link to this extract
When you try to buy online ads from Facebook’s self-serve ad-auctioning platform, merely being the highest bidder isn’t enough to guarantee that your ads will get through: Facebook multiplies your bid by a software-generated prediction about how responsive the audience will be to it, so the clickbaitier your ad is, the less it costs to place it.
This is just one of the insights into the odd attention economics lessons in Antonio García Martínez’s deep dive into the Trump campaign’s use of Facebook in the 2016 election; another important lesson is that poor, rural voters aren’t very attractive to advertisers, so there’s less competition when you try to reach them and that makes them cheaper to advertise to than voters in big, Democratic-leaning cities.
The really interesting stuff is about the six-year-old Facebook practice of selling ads to Custom Audiences (people who engaged in a specific activity that Facebook knows about, like putting a pair of shoes in an online shopping basket, visiting a given website, etc), and Lookalike Audiences (people who have similar characteristics to a Custom Audience, that is, “Advertise this to people substantially similar to people you know about who have recently gone shoe-shopping”).
»Unsurprisingly, the Russians also apparently made use of Custom Audiences in their ads campaign. The unwary clicker on a Russian ad who then visited their propaganda site suddenly could find yet more planted content in their Feed, which could generate downstream engagement in Feed, and thus the great Facebook wheel turned. The scale of their spend was puny, however, a measly $100,000, which pales in comparison to the millions Trump spent on online advertising.
The above isn’t mere informed speculation, the Trump campaign admitted to its wide use of both Custom and Lookalike audiences. There seems to be little public coverage of whether the Clinton campaign used Facebook Ads extensively, but there’s no reason to think her campaign did not exploit the same tools.«
There was a lot of surprise about this when I tweeted an extract from the Wired piece (which I can’t get to load; it’s linked in this piece) on Saturday. That’s because we don’t expect ads to work like that. We’re familiar with a straight transaction – spend X, show to Y people. The idea that the platform itself inserts a multiplier seems weird to anyone unversed in it.
And it also implies that you can do better with crazy just-about lies than with calm, reasoned statements about what you’ll do. That doesn’t imply good things about closing the US partisan gap.
link to this extract
Amazon AWS S3 cloud storage servers might soon fall victims to ransom attacks, similar to how hacker groups held tens of thousands of MongoDB databases for ransom throughout 2017.
The statement, made today on social media by infosec expert Kevin Beaumont, is nothing short of a prophecy of things to come, an opinion shared by many security professionals to whom Bleeping Computer spoke today.
Amazon AWS S3 storage servers have been leaking data all 2017, being behind some of the most notable data leaks of last year, including breaches at the NSA, the US Army, analytics providers, and more.
Those incidents happened because companies left data on publicly-readable S3 buckets (“bucket” being a term used to describe an S3 storage unit). In most cases, that data was found by security researchers who helped companies secure their systems, but hackers could get to these files first, too.
However, there’s also a category of S3 buckets that are even more dangerous than publicly-readable servers. Those are publicly-writeable ones —buckets allowing any user, with or without an Amazon S3 account, to write or delete data on the AWS S3 instance. A Skyhigh Networks report from September 2017 found that 7% of all Amazon AWS S3 buckets were publicly-writeable.
Experts believe that hacker groups who have been busy holding MongoDB, ElasticSearch, Hadoop, CouchDB, Cassandra, and MySQL servers for ransom all of 2017 might soon turn their sights on S3 publicly-writeable buckets.
The 2017 ransom attacks usually followed the same pattern. Hackers found an exposed server, wiped data, and left a ransom note behind asking for a ransom. Some victims paid, hoping to recover data, but most users were left at the altar, as hackers did not have the storage space to back up all the ransomed servers, and never returned any of the promised data.
Now, something like this is bound to happen to Amazon S3 server owners.
Basically, open season for any capable hacker to take over.
link to this extract
When Apple begins hosting Chinese users’ iCloud accounts in a new Chinese data center at the end of this month to comply with new laws there, Chinese authorities will have far easier access to text messages, email and other data stored in the cloud.
That’s because of a change to how the company handles the cryptographic keys needed to unlock an iCloud account. Until now, such keys have always been stored in the United States, meaning that any government or law enforcement authority seeking access to a Chinese iCloud account needed to go through the US legal system.
Now, according to Apple, for the first time the company will store the keys for Chinese iCloud accounts in China itself. That means Chinese authorities will no longer have to use the US courts to seek information on iCloud users and can instead use their own legal system to ask Apple to hand over iCloud data for Chinese users, legal experts said.
Human rights activists say they fear the authorities could use that power to track down dissidents, citing cases from more than a decade ago in which Yahoo handed over user data that led to arrests and prison sentences for two democracy advocates. Jing Zhao, a human rights activist and Apple shareholder, said he could envisage worse human rights issues arising from Apple handing over iCloud data than occurred in the Yahoo case.
In a statement, Apple said it had to comply with recently introduced Chinese laws that require cloud services offered to Chinese citizens be operated by Chinese companies and that the data be stored in China. It said that while the company’s values don’t change in different parts of the world, it is subject to each country’s laws.
China today – and Russia, Turkey, the Philippines tomorrow, if they pass similar laws? Where does it stop? An iCloud backup also includes iMessage, so this is a risk to activists. I expect they will take two countermeasures: stop using iCloud (and delete all their backups), and start (or continue) using apps such as Signal.
It’s a sign of the roach motel effect of China on Apple: it’s such a big slice of its business now that it can’t (unlike Google in 2010) just refuse to do business there.
link to this extract
Manafort left an incriminating paper trail because he couldn’t figure out how to convert PDFs to Word files • Slate
So here’s the essence of what went wrong for [Trump aides] Manafort and Gates, according to Mueller’s investigation: Manafort allegedly wanted to falsify his company’s income, but he couldn’t figure out how to edit the PDF. He therefore had Gates turn it into a Microsoft Word document for him, which led the two to bounce the documents back-and-forth over email. As attorney and blogger Susan Simpson notes on Twitter, Manafort’s inability to complete a basic task on his own seems to have effectively “created an incriminating paper trail.”
In Manafort’s defense, converting documents to and from Word could be easier. Not having tried it for a while, I attempted to transform my Word draft of this blog post into a PDF. I confess that I did fumble a bit at first (it’s been a while), but I eventually managed to get the job done. According to my stopwatch, the full ordeal took me 42 seconds. It involves a few steps, but there are plenty of accessible tutorials out there if you get lost.
Changing PDFs back to editable Word documents, meanwhile, does get a little more complicated. Try it in Adobe Acrobat (via the “Save as Other” command under “File” on a Mac) and you’ll quickly be redirected to Adobe’s website and presented with a handful of subscription packages that will allow you to transform your documents. For as little as $2 a month, Adobe will allow you to convert PDF files to Word, Excel, and rich text formats. If this feels extortionate, there are also plenty of services online that promise to let you do the same thing for free, but—and, to be clear, I’m no financial genius—even people who are allegedly misreporting millions of dollars in income can almost certainly afford the budget option. Indeed, it’s probably a little safer, all things considered.
This is a little like the Mars lander which crashed because one team used imperial units and the other used metric. Tiny technical details bringing down a huge enterprise.
link to this extract
Launched earlier this week, the Petro raised $735m in its first day, despite investor warnings from groups including the US Treasury Department. Maduro has said he plans to offer a total of 100 million Petros, with a starting price of $60 each, for a total of about $6bn.
“Our country has released our first official crypto in the history of the world,” Maduro said in a nationally televised event. “It’s also the only one whose value is backed by real estate. The Petro demonstrates, today more than ever, that together all is possible.”
In making the proposal, Maduro added that banks that build mining farms could increase the benefits of their employees. Union leaders, though, rejected the proposal.
“That seems to us an abuse of power and a totalitarianism,” said Ana Yanez, the national coordinator of the National Union of Workers. “In addition, [the Petro] is a virtual currency that violates the Constitution. As workers, we disagree that this cryptocurrency is imposed on us.”
Oh, but that’s not the end of it. The catch to the Petro is how difficult it is to redeem: you can’t redeem it for actual oil. You redeem it for the value of a barrel of oil, paid for in bolivars – whose own value is dwindling at a huge rate.
link to this extract
Ai facial recognition works better for white skin – because it’s being trained that way • World Economic Forum
Three commercially released facial-analysis programs from major technology companies demonstrate both skin-type and gender biases, according to a new paper researchers from MIT and Stanford University will present later this month at the Conference on Fairness, Accountability, and Transparency.
In the researchers’ experiments, the three programs’ error rates in determining the gender of light-skinned men were never worse than 0.8%. For darker-skinned women, however, the error rates ballooned — to more than 20% in one case and more than 34% in the other two.
The findings raise questions about how today’s neural networks, which learn to perform computational tasks by looking for patterns in huge data sets, are trained and evaluated. For instance, according to the paper, researchers at a major US technology company claimed an accuracy rate of more than 97% for a face-recognition system they’d designed. But the data set used to assess its performance was more than 77% male and more than 83% white.
“What’s really important here is the method and how that method applies to other applications,” says Joy Buolamwini, a researcher in the MIT Media Lab’s Civic Media group and first author on the new paper. “The same data-centric techniques that can be used to try to determine somebody’s gender are also used to identify a person when you’re looking for a criminal suspect or to unlock your phone. And it’s not just about computer vision. I’m really hopeful that this will spur more work into looking at [other] disparities.”
Would love to know which big American company that was.
link to this extract
Given enough time, everything withers and dies, from the most robust institutions to the most popular crowdsales. No one expected all of 2017’s ICOs to last the course. The pace at which they’ve withered and died may come as a surprise though. Tokendata, one of the more comprehensive ICO trackers, lists 902 crowdsales which took place last year. Of these, 142 failed at the funding stage and a further 276 have since failed, either due to taking the money and running, or slowly fading into obscurity. This means that 46% of last year’s ICOs have already failed.
The number of ICOs that are still a going concern is actually even lower. An additional 113 ICOs can be classified as “semi-failed”, either because their team has stopped communicating on social media, or because their community is so small as to mean the project has no chance of success. This means that 59% of last year’s crowdsales are either confirmed failures or failures-in-the-making.
Are social platforms changing the way we judge art, or making us more easily dismissive? Put another way, does a work that an artist has labored over for months, even years, deserve more than a glance on a tiny screen?
As the images of the Obama portraits started streaming out, the reactions came hard and fast. Looking at art online feeds a coterie of casual critics, just as everyone’s a design critic with every new logo launch these days. Within minutes of the live-streamed unveiling, Twitter was ablaze with hot takes, critiques and memes.
A seminal 2001 study found that museum goers spent an average of 17 seconds looking at a work of art in a museum, with the bulk of time spent the reading the wall text. With social media, this time is probably even shorter.
There was a time when art lovers traveled great distances to see a work of art. For these pilgrims, a personal audience with genius was the goal. A museum was a place to hone one’s ability to detect beauty and appreciate nuance—with only our own internal filters between us and the work.
Platforms like the incredible Google virtual museum tours have eliminated the need to travel, by enabling art lovers to “visit” top museums and “see” works of art up close, at even higher resolutions than if they were to stand before it a museum.
But despite the democratizing value of widely disseminating great masterpieces, the fact is that looking at art on our backlit screens is not the same as encountering it in person. Take the work of the British painter J.M.W. Turner: To the casual Instagram swiper, his wild brushstrokes might seem unruly, even quaint. In person, even a smaller canvas like Peace – Burial at Sea is so arresting and emotional it’s impossible to ignore.
I’m probably the last person anyone would call an art connoisseur, but the reality is that art, including paintings, is three-dimensional. (Sculpture, of course, even more so.) The swirls of paint on a van Gogh or Pisarro tell you about the artist’s intent and skill. A photo – even a virtual museum – won’t show you that.
link to this extract
China’s cutthroat smartphone market is coming down to a handful of major brands • South China Morning Post
“The top-20 smartphone brands control 93% of the market,” said Counterpart research director Neil Shah. “That means 180 other brands are competing for just the remaining seven% share, which means we could see a potential exit for some of these firms this year.”
He said the slowdown [in smartphone sales, which went into reverse in Q4 17] in China “has caused serious pain for tier-2 and tier-3 smartphone brands, such as Gionee, Coolpad and LeEco, which largely depend on domestic sales”. A further slowdown this year would make it more difficult for these companies to compete in terms of scale in component supply deals for production and in marketing, he said.
With fewer marketing resources at their disposal, small Chinese smartphone companies will be absent at this year’s edition of Mobile World Congress (MWC), the world’s largest exhibition for the mobile industry, to be held in Barcelona, Spain, from 26 February to 1 March.
This year’s absentees include Meizu, Gionee and Coolpad, which took part the past three years, and LeEco, which was at the event in 2015 and 2016.
Meizu insists it’s profitable, but I think the others face a cold winter.
link to this extract
Errata, corrigenda and ai no corrida: none notified.