Start Up: Cambridge Analytica in hot water, Alexa’s deep (AI) music, YouTube Kids joins the info wars, and more

Mattress reviews online are replacing sales people in stores. But not for the better. Photo by Sameer Vasta on Flickr.

You can sign up to receive each day’s Start Up post by email. You’ll need to click a confirmation link, so no spam.

A selection of 13 links for you. “I’m.. working on a blockchain gang”. I’m @charlesarthur on Twitter. Observations and links welcome.

DeepMusic Alexa skill serves up AI-generated songs • MusicAlly

Stuart Dredge:


Amid all the industry conversation about how smart speakers will affect the way people listen to music, the assumption has been that the music in question will be made by humans.

Here’s a new Alexa skill to make you think, though. It’s called DeepMusic, and has just launched for Alexa-powered devices like the Echo speakers.

“DeepMusic is an Alexa skill that enables you to listen to songs generated by artificial intelligence (AI). Each song was composed entirely using AI. The songs were generated using a collection of audio samples and a deep recurrent neural network. There has been no post-production editing by a human,” explains its description on Amazon’s store.

AI was also used to create the artwork shown on the screen-equipped Echo Show and Echo Spot speakers. The skill can be tested by saying ‘Alexa, open DeepMusic’ and then commands like ‘Alexa, ask DeepMusic to play a song’.


We’ve had quite a few “AI music” links over the past few years. There was in August 2016, an AI-generated song in November 2016, and DeepBach in December 2016. If anyone wants to let us know how DeepMusic sounds, we’d love a review.
link to this extract

To find suspects, Raleigh police quietly turn to Google •

Tyler Dukes, WRAL public records reporter:


In the early morning hours of June 1, 2015, witnesses heard Pugh in an argument. They heard multiple gunshots. And they saw a figure illuminate the ground with a cell phone flashlight before fleeing the scene as the sun rose over north Raleigh.

A year and a half later across town, a taxi driver named Nwabu Efobi was gunned down in front of the Universal Cab Company. Security camera video caught Efobi in some kind of confrontation with the shooter before the unknown man opened fire. The day before, cameras caught the same guy several times walking around the building with what appeared to be a cell phone at his ear.

Raleigh police say the cases are unrelated. But in March 2017, months after investigations began into both shootings, separate detectives on each case, one day apart, employed an innovative strategy in criminal investigations.

On a satellite image, they drew shapes around the crime scenes, marking the coordinates on the map. Then they convinced a Wake County judge they had enough probable cause to order Google to hand over account identifiers on every single cell phone that crossed the digital cordon during certain times.

In at least four investigations last year – cases of murder, sexual battery and even possible arson at the massive downtown fire in March 2017 – Raleigh police used search warrants to demand Google accounts not of specific suspects, but from any mobile devices that veered too close to the scene of a crime, according to a WRAL News review of court records.


Two things: first, the idea of a “public records reporter” wouldn’t exist in the UK; they’re either incredibly hard to access, or trivial. Second, why only Google? What if there were iPhone users who didn’t use Google services in the area?
link to this extract

The online mattress industry promised a more honest shopping experience. We may have gotten the opposite instead • Recode

Jason Del Rey:


Gone is the slick-talking store sales guy directing you toward the mattress with the best profit margin or the highest commission. But in his place are a slew of website owners proficient in the art of search engine marketing, funneling you with the help of Google toward the mattress that lines their pockets the most.

These sites make money from so-called affiliate fees — commissions earned when a reader clicks a link in a review and goes on to purchase that mattress afterward. This model has been around almost as long as the internet. But the rise of online mattress sellers has created a perfect recipe for these content chefs: a high-priced item that results in a large commission, coupled with a heavy consumer reliance on reviews, since many of these new mattress brands are not widely sold in physical stores.

The power these websites amassed has not gone unnoticed in the industry. After Casper sued three of the sites, the high-profile mattress company financed the takeover of one of them, called Sleepopolis. This raised questions about a conflict of interest when Sleepopolis’ review of Casper suddenly improved. Today, Sleepopolis sends more traffic to Casper’s website than to any other mattress brand, according to data from SimilarWeb.


You might recall this amazing story about Sleepopolis and Casper and the whole crazy setup from October 2017. Another example of “the internet will improve it all” not coming true.
link to this extract

A cyberattack in Saudi Arabia had a deadly goal. Experts fear another try • The New York Times

Nicole Perlroth and Clifford Krauss:


In August, a petrochemical company with a plant in Saudi Arabia was hit by a new kind of cyberassault. The attack was not designed to simply destroy data or shut down the plant, investigators believe. It was meant to sabotage the firm’s operations and trigger an explosion.

The attack was a dangerous escalation in international hacking, as faceless enemies demonstrated both the drive and the ability to inflict serious physical damage. And United States government officials, their allies and cybersecurity researchers worry that the culprits could replicate it in other countries, since thousands of industrial plants all over the world rely on the same American-engineered computer systems that were compromised.

Investigators have been tight-lipped about the August attack. They still won’t identify the company or the country where it is based and have not identified the culprits.

But the attackers were sophisticated and had plenty of time and resources, an indication that they were most likely supported by a government, according to more than a dozen people, including cybersecurity experts who have looked into the attack and asked not to be identified because of the confidentiality of the continuing investigation.

The only thing that prevented an explosion was a mistake in the attackers’ computer code, the investigators said…

…What worries investigators and intelligence analysts the most is that the attackers compromised Schneider’s Triconex controllers, which keep equipment operating safely by performing tasks like regulating voltage, pressure and temperatures. Those controllers are used in about 18,000 plants around the world, including nuclear and water treatment facilities, oil and gas refineries, and chemical plants.

“If attackers developed a technique against Schneider equipment in Saudi Arabia, they could very well deploy the same technique here in the United States,” said James A. Lewis, a cybersecurity expert at the Center for Strategic and International Studies, a Washington think tank.


Most likely to be Iranian hackers. Second-tier nation-state hackers are now as big a problem as the top tier; they just don’t hit as many targets.

link to this extract

University of Arizona tracks student ID cards to detect who might drop out • The Verge

Shannon Liao:


The University of Arizona is tracking freshman students’ ID card swipes to anticipate which students are more likely to drop out. University researchers hope to use the data to lower dropout rates. (Dropping out refers to those who have left higher-education entirely and those who transfer to other colleges.)

The card data tells researchers how frequently a student has entered a residence hall, library, and the student recreation center, which includes a salon, convenience store, mail room, and movie theater. The cards are also used for buying vending machine snacks and more, putting the total number of locations near 700. There’s a sensor embedded in the CatCard student IDs, which are given to every student attending the university.

“By getting their digital traces, you can explore their patterns of movement, behavior and interactions, and that tells you a great deal about them,” Sudha Ram, a professor of management information systems who directs the initiative, said in a press release.

Researchers have gathered freshman data over a three-year time frame so far, and they found that their predictions for who is more likely to drop out are 73% accurate.


Big data brother is everywhere.
link to this extract

‘I created Steve Bannon’s psychological warfare tool’: meet the data war whistleblower • The Observer

Carole Cadwalldr:


What the email correspondence between Cambridge Analytica employees and [Aleksandr] Kogan shows is that Kogan had collected millions of profiles in a matter of weeks. But neither Wylie nor anyone else at Cambridge Analytica had checked that it was legal. It certainly wasn’t authorised. Kogan did have permission to pull Facebook data, but for academic purposes only. What’s more, under British data protection laws, it’s illegal for personal data to be sold to a third party without consent.

“Facebook could see it was happening,” says Wylie. “Their security protocols were triggered because Kogan’s apps were pulling this enormous amount of data, but apparently Kogan told them it was for academic use. So they were like, ‘Fine’.”

Kogan maintains that everything he did was legal and he had a “close working relationship” with Facebook, which had granted him permission for his apps.

Cambridge Analytica had its data. This was the foundation of everything it did next – how it extracted psychological insights from the “seeders” and then built an algorithm to profile millions more.

For more than a year, the reporting around what Cambridge Analytica did or didn’t do for Trump has revolved around the question of “psychographics”, but Wylie points out: “Everything was built on the back of that data. The models, the algorithm. Everything. Why wouldn’t you use it in your biggest campaign ever?”

In December 2015, the Guardian’s Harry Davies published the first report about Cambridge Analytica acquiring Facebook data and using it to support Ted Cruz in his campaign to be the US Republican candidate. But it wasn’t until many months later that Facebook took action. And then, all they did was write a letter. In August 2016, shortly before the US election, and two years after the breach took place, Facebook’s lawyers wrote to [data whistleblower Christopher] Wylie, who left Cambridge Analytica in 2014, and told him the data had been illicitly obtained and that “GSR was not authorised to share or sell it”. They said it must be deleted immediately.

“I already had. But literally all I had to do was tick a box and sign it and send it back, and that was it,” says Wylie. “Facebook made zero effort to get the data back.”


Facebook staff and ex-staff have been saying: 1) not a data breach as such (a headline on another of these stories) 2) this stuff doesn’t affect anyone anyway. The second isn’t believable. Are we meant to think advertising has no effect? If that were the case, nobody would do it.
link to this extract

Followup questions For Facebook, Cambridge Analytica and Trump Campaign on massive breach • Just Security

Justin Hendrix:


Journalists, regulatory bodies and Congress should be ready to ask a number of pressing questions to get to the bottom of exactly what happened. The answers are important- governments around the world are considering how best to regulate technology companies, and this extraordinary incident gets to the heart of the relationship between personal data, microtargeting, dark money and the impact of their combination with unaccountable platforms on the health of democracies.

Here are seven key questions:
[with the explanation removed, here are the questions:]
1. Why did Facebook take more than two years to inform the public of this massive breach?
2. Did the Trump campaign or Cambridge Analytica violate campaign finance laws?
3. Did Trump campaign or Cambridge Analytica employees lie to Congress, or to the British Parliament?
4. Did Facebook’s failure to disclose this breach to the public and notify its directly affected consumers break any laws?
5. Did any of the Facebook embeds in the Trump campaign know that stolen data was being used for targeting?
6. Did Facebook have evidence its own employees mishandled this situation? Was any disciplinary action taken?
7. Did other organizations or individuals exploit these apparent weaknesses, and are there other breaches we do not know about?


Question 3 is already being followed up.
link to this extract

Facebook bans Trump campaign’s data analytics firm for taking user data • The Washington Post

Elizabeth Dwoskin:


Facebook said Kogan had requested and gained access to information from 270,000 Facebook members after they chose to download his app. The app, “thisisyourdigitallife,” offered a personality prediction and billed itself on Facebook as “a research app used by psychologists.”

The Facebook members gave their consent for Kogan to access information such as the city they set on their profile, the content they had liked and some limited information about friend groups and contacts. Kogan then broke Facebook’s policies and passed the information to Cambridge Analytica and to Wylie. Facebook learned about Kogan’s activities in 2015.

The company removed Kogan’s app at the time and demanded certifications from Cambridge Analytica, Wylie and Kogan that the information he had shared had been destroyed. All three certified to Facebook that they had done so, but Facebook said it received reports several days ago that the data was not deleted.


“The horse bolted two years ago but to indicate our displeasure we are definitely shutting this stable door.” Also amusing: where the reports came from. It then threatened to sue one of the sources (The Observer) of those reports.
link to this extract

YouTube suggested conspiracy videos to children using its Kids app • Business Insider

James Cook:


YouTube’s app specifically for children is meant to filter out adult content and provide a “world of learning and fun,” but Business Insider found that YouTube Kids featured many conspiracy theory videos which make claims that the world is flat, that the moon landing was faked, and that the planet is ruled by reptile-human hybrids.

YouTube Kids is a separate app from the main YouTube app, and it’s meant to allow parents to let their children browse YouTube without being worried about any unsuitable content appearing. Children are encouraged to learn languages, read books, and watch educational videos.

Search for “UFO” on YouTube Kids and you’ll mostly find videos of toys that are clearly fine for children to watch. But one of the top videos claimed to show a UFO shooting at a chemtrail, and we found several videos by prominent conspiracy theorist David Icke in the suggested videos. YouTube removed the videos from YouTube Kids after we contacted it about the issue.

One suggested video was an hours-long lecture by Icke in which he claims that aliens built the pyramids, that the planet is run by reptile-human hybrids, that Freemasons engage in human sacrifice, that the assassination of President Kennedy was planned by the US government, and that humans would evolve in 2012.


With enough examples, is it possible that Google will actually take action before rather than after this happens? Except it’s been evident for ages that YouTube Kids is a complete mess. This just shows that it’s even worse than all those algorithmically-generated junk weird videos.

And boring though it might be to repeat this, kids brought up on the Encyclopaedia Britannica didn’t come across David Icke or moon landing conspiracies.
link to this extract

A quick (battle) field guide to the new culture wars • Ribbonfarm

Venkatesh Rao:


I haven’t yet gathered my thoughts on this whole theme properly, besides drawing my map, but I did tweet a bunch yesterday (March 5), throwing out some initial thoughts on the culture wars topic. I’m compiling them here.

The overarching theme of these tweets is this: there is a war on, and except for the low level of actual killing, it is a real war, not an allegorical or metaphoric one. The most visible battlefields are online forums like Twitter, Facebook, and various well-connected regions of the blogosphere. But there is also plenty of old-school direct action on the streets, in traditional media outlets, and behind closed doors.

The combatants include professional cyberwarriors and seasoned amateur guerrillas pursuing very well-defined objectives with military precision and specialized tools. Then there is the small but highly skilled corps of shitposters whose skill at information warfare is matched only by their fundamental incomprehension of the real damage they’re unleashing for lulz. And finally, masses of clueless patsies being programmed like insect swarms by all sides. What Renee DiResta labeled always-on mobs in her post last year.

In other words, there is a war on, it’s very real, causing real pain to many, and involves huge consequences hanging in the balance, from the future of academia and the conduct of science to the future of the planet itself.

Believe it or not, the swinging of a presidential election is actually a fairly minor chapter in the ongoing saga. When it’s all done and over with, and the dust has settled somewhat, I believe we’ll look back on this era as being as consequential in reshaping the future of the United States and the world as the Civil War.


This is a deeply fascinating post by Rao, and really obligatory reading if you are on social media at all. His comment on conspiracy theorists – they aren’t going to go away, and you might as well treat them as being from a parallel universe (I paraphrase) is worth the time in itself.

It’s also a concerning post. But knowing you’re in the midst of an infowar is perhaps the most useful information you can have.
link to this extract

Beyond the bitcoin bubble, the benefits of blockchain • Allianz Global Investors

Stefan Hofrichter is head of global economics and strategy at Allianz:


bitcoin ticks all of the boxes that we consider to be essential criteria of any asset bubble:

• “New-era” thinking. Bitcoin is perceived to be an entirely new kind of currency and a monetary innovation in the internet age.
• Overtrading. Trading volumes have increased by almost fivefold in the last five years, according to BIS data.
• Ultra-easy monetary conditions. Accommodative policy is still in place globally, despite a series of rate hikes by the US Federal Reserve.
• A lack of financial regulation. The “Wild West” bitcoin environment is only gradually being addressed by regulators around the world.
• The launch of related financial instruments. New products related to the bubbling asset class are popping up – from CBOE and CME futures contracts to the launch of “ICOs” (initial coin offerings).
• Rising leverage. Not only has private-sector leverage increased to record highs globally, but leveraged speculation in bitcoin is increasing.
• Swindles. Bitcoin has become the instrument of choice for many criminals, thanks to its ability to exist entirely outside of traditional banking channels.
• Significant overvaluation. Many other asset classes are pricey in today’s market, but bitcoin’s valuation seems to be without peer.

This brings us to a key question: what is the fair value of a bitcoin? In our view, its intrinsic value must be zero: a bitcoin is a claim on nobody – in contrast to, for instance, sovereign bonds, equities or paper money – and it does not generate any income stream.


Well this won’t be popular with the bitcoin miners.
link to this extract

Sierra Leone tests blockchain technology for tallying election results •

Daniel Finnan:


Sierra Leone has become one of the first countries to trial the use of blockchain technology during elections. A Swiss-based company called Agora used technology similar to that underpinning the BitCoin cryptocurrency to provide a partial count of election results. Sierra Leone went to the polls on 7 March to elect a new president, members of parliament, local councillors and mayors.

“We recorded the votes after tallying on our blockchain where the votes are anonymised,” Leonardo Gammar, the Chief Executive Officer of Agora, told RFI. “Our observers have a look at the ballots and they send the results on our blockchain per polling station,” he added.

The Agora e-voting system aims to provide a decentralised system that is both transparent and verifiable, according to Gammar. Blockchain technology emerged with the creation of Bitcoin and is a digital ledger of records much like a traditional database, except that it is encrypted and stored across several computers.

“Our goal is to provide voting solutions for people, electronic voting solutions, but decentralised,” said Gammar, during a telephone interview. “In Sierra Leone, what we did is just a use case and it’s not the full implementation of our digital solution.”

Agora aims to provide a full e-voting system from voter identification to the actual casting of ballots and counting of results in the future, Gammar said.

Agora’s voting app will be linked to this implementation of blockchain technology and will provide additional features, the Agora co-founder said. One such feature would help avoid electoral fraud such as vote buying. The app would enable voters to record their choice in the election when under pressure from unscrupulous actors, but change the vote to their desired choice at a later time, as long as the poll is still open.


That last bit is clear as mud. Also: if people are being pressured to sell their votes and there’s a public tally of how many votes there were for something.. that doesn’t seem good? And it doesn’t seem as though it changes the circumstance that existed before: either the vote-buyer could come into the voting booth and watch, or they had to take it on trust. (This is e-voting, so the risk of vote-buying being enforced seems bigger than if you have voting booths.)

This seems like a buzzword solution; encrypting the ledger and storing it on multiple computers is what any organisation that handles distributed logins does for its passwords. It does, though, have the benefit that there’s a hard-to-dispute distributed tally of recorded votes.
link to this extract

Truth isn’t the problem—we are • WSJ

Rebecca Newberger Goldstein on how “post-truth” is more about allegiance than truth:


Consider two different propositions, from opposite ends of American politics: (1) “The only way to stop violent crime is to allow citizens to arm themselves,” and (2) “For a person of privilege to make creative use of the culture of the underprivileged is an act of aggression and abuse.” The information that we can glean from these statements isn’t about the putative topics—gun control or cultural “appropriation,” respectively. It’s about the political identity of the speaker. Such assertions are tribal banners, and offering counter-evidence isn’t likely to get you very far.

Indeed, a pledge of political allegiance achieves greater authenticity if it flies in the face of counter-evidence, especially if that evidence comes from “so-called experts.” My insistence that “Human actions have no impact on global warming” gains immeasurably, as a pledge, from the fact that 97% of climate scientists disagree with me; it highlights the depth of my commitment to the cause. Similarly, to show my solidarity with others who wish to ban “Frankenstein” foods, I can insist that “Genetically engineered crops are unsafe for humans and animals,” even as I’m presented with an exhaustive study by the National Academies of Science concluding that there is no such evidence.

These pseudo-assertions aren’t just tribal markers, of course. They also purport to say what is and isn’t true. And that’s where we get into trouble—in the very fact that their persuasive potency, as pledges, is often a function of how far they depart from the best available evidence.

Democratic debate is never a strict weighing of evidence; emotional appeals to party, cause and country are always part of the mix. But our readiness today to proudly defy evidence is very troubling. It undermines our commitment to the truth—and our capacity to reach any sort of middle ground or consensus.


It generally seems to me that the problem in the US is the lack of a middle ground in politics. One counter to that has been the recent upsets in which Democrats have won with huge poll swings – but they’re in effect obliged to be middle-of-the-road to appeal to potential swing voters. If they can hang on, America might find itself inching back toward consensus through democratic (small d) means.
link to this extract

Errata, corrigenda and ai no corrida: none notified

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.