Ever finish a book and think “This would make a great role-playing game!”? Me too! My shelves are filled with role-playing games based on various books¹. Some were successful adaptations. Others, not so much. Having spent seventeen years selling RPGs, I have some ideas about what sort of stories adapt well to games and which […]

These Spam snatchers are not hungry people desperate for Spam, said Tina Yamaki, president of the Retail Merchants of Hawaii. They are most likely part of a Spam black market that’s taking off in a state where the demand for Spam knows no bounds.

When I give the dating app LoveFlutter my Twitter handle, it rewards me with a 28-axis breakdown of my personality: I’m an analytic Type A who’s unsettlingly sex-focused and neurotic (99th percentile). On the sidebar where my “Personality Snapshot” is broken down in further detail, a section called “Chat-Up Advice” advises, “Do your best to avoid being negative. Get to the point quickly and don’t waste their time. They may get impatient if you’re moving too slowly.” I’m a catch.

Loveflutter, a Twitter-themed dating app from the UK, doesn’t ask me to fill out a personality survey or lengthy About Me (it caps my self-description at a cute 140 characters). Instead, it’s paired with the language processing company Receptiviti.ai to compute the compatibility between me and its user base using the contents of our Twitter feeds. Is this good matchmaking or a gimmick? As a sex-crazed neurotic, I think you know where I stand.

Dating apps promise to connect us with people we’re supposed to be with—momentarily, or more—allegedly better than we know ourselves. Sometimes it works out, sometimes it doesn’t. But as machine learning algorithms become more accurate and accessible than ever, dating companies will be able to learn more precisely who we are and who we “should” go on dates with. How we date online is about to change. The future is brutal and we’re halfway there.

“Personality” surveys

Today, dating companies fall into two camps: sites like eHarmony, Match, and OkCupid ask users to fill out long personal essays and answer personality questionnaires which they use to pair members by compatibility (though when it comes to predicting attraction, researchers find these surveys dubious). Profiles like these are rich in information, but they take time to fill out and give daters ample incentive to misrepresent themselves (by asking questions like, “How often do you work out?” or “Are you messy?”). On the other hand, companies like Tinder, Bumble, and Hinge skip surveys and long essays, instead asking users to link their social media accounts. Tinder populates profiles with Spotify artists, Facebook friends and likes, and Instagram photos. Instead of matching users by “compatibility,” these apps work to provide a stream of warm bodies as fast as possible.

It’s true that we reveal more of ourselves in Twitter posts, Facebook likes, Instagram photos, and Foursquare check-ins than we realize. We give dating apps access to this data and more: when one journalist from The Guardian asked Tinder for all the information it had on her, the company sent her a report 800 pages long. Sound creepy? Maybe. But when I worked as an engineer and data scientist at OkCupid, massive streams of data like these made me drool.

In the future, apps like Tinder may be able to infer more about our personalities and lifestyles through our social media activity than an eHarmony questionnaire ever could capture. Researchers already think they can predict how neurotic we are from our Foursquare check-ins, whether or not we’re depressed from our Tweets and the filters we choose on Instagram, and how intelligent, happy, and likely to use drugs we are from our Facebook likes.

What’s more, the relationship between our online behavior and what it implies about us is often unintuitive. One 2013 study from Cambridge University that analyzed the connection between Facebook likes and personality traits found the biggest predictors of intelligence were liking “Science” and “The Colbert Report” (unsurprising) but also “Thunderstorms” and “Curly Fries.” That connection might defy human logic, but what does that matter if you’re feeding a personality algorithm into a matchmaking algorithm?

Social media sousveillance

Because indicators of our personality can be subtle, and we tend not to curate our activity on Facebook as closely as we might a dating profile, perhaps there’s more integrity to this data than what users volunteer in survey questions.

“My initial reaction to online dating is that people might present a version that’s unrealistic,” said Chris Danforth, Flint professor of Mathematical, Natural, and Technical Sciences at the University of Vermont who’s studied the link between Instagram, Twitter, and depression. “But what seems to be revealed every time one of these studies comes out is that it looks to be the case that we reveal more about ourselves than we realize, maybe not as much in solicited surveys but in what we do. Someone’s likes on Facebook could be a better predictor of whether they would get along with someone than survey answers.”

The data could also be used to keep users honest when they’re making their accounts. “I think it would be interesting if OkCupid called you out as you’re filling out your profile,” said Jen Golbeck, a researcher who studies the intersection of social media and information at the University of Maryland. “It could say something like, ‘I analyzed your likes and it looks like maybe you are a smoker. Are you sure you want to choose that answer?’” A more jaded dating app could instead alert the person viewing the profile that their match might be lying.

Companies could use insights from daters’ online behavior to catch red flags and prevent some people from joining in the first place. After the Charlottesville white nationalist rally in August, some dating services asked members to report white supremacists and banned them. But in the future, apps could identify sexists/racists/homophobes by their social media activity and preemptively blacklist them from joining. (Maybe this would aid the industry’s problem with harassment, too.)

But they could also ban users who display personality traits that allegedly don’t work well in relationships. eHarmony, for example, rejects applicants who’ve been married four or more times, or, in an ableist twist, those whose survey responses indicate they might be depressed. A dystopian future dating algorithm could flag users who are depressed or suffering from anxiety from their posts, likes or Tweets, and reject them.

Algorithms could also use our online behavior to learn the real answers to questions we might lie about in a dating questionnaire. One of OkCupid’s matching questions, for example, asks “Do you work out a lot?” But MeetMeOutside, a dating app for sporty people, asks users to link their Fitbits and prove they’re physically active through their step counts. This type of data is harder to fake. Or, rather than ask someone whether they’re more likely to go out or Netflix and chill on a Friday night, a dating app could simply collect this data from our GPS or Foursquare activity and pair equally active users.

The algorithm religion

It’s also possible that computers, with access to more data and processing power than any human, could pick up on patterns human beings miss or can’t even recognize. “When you’re looking through the feed of someone you’re considering, you only have access to their behavior,” Danforth says. “But an algorithm would have access to the differences between their behavior and a million other people’s. There are instincts that you have looking through someone’s feed that might be difficult to quantify, and there may be other dimension we don’t see… nonlinear combinations which aren’t easy to explain.”

Just as dating algorithms will get better at learning who we are, they’ll also get better at learning who we like—without ever asking our preferences. Already, some apps do this by learning patterns in who we left and right swipe on, the same way Netflix makes recommendations from the movies we’ve liked in the past. 

“Instead of asking questions about individuals, we work purely on their behavior as they navigate through a dating site,” says Gavin Potter, founder of RecSys, a company whose algorithms power tens of niche dating apps. “Rather than ask someone, ‘What sort of people do you prefer? Ages 50-60?’ we look at who he’s looking at. If it’s 25-year-old blondes, our system starts recommending him 25-year-old blondes.” OkCupid data shows that straight male users tend to message women significantly younger than the age they say they’re looking for, so making recommendations based on behavior rather than self-reported preference is likely more accurate.

Algorithms that analyze user behavior can also identify subtle, surprising, or hard-to-describe patterns in what we find attractive—the ineffable features that make up one’s “type.” Or at least, some app makers seem to think so.

“If you look at the recommendations we generated for individuals, you’ll see they all reflect the same type of person—all brunettes, blondes, of a certain age,” Potter says. “There are women in Houston who only want to go out with men with beards or facial hair. We found in China users who like a very, um, demure type of individual.” This he mentions in a tone which seems to imply a stereotype I’m unaware of. “No questionnaire I’m aware of captures that.”

Naturally, we might not like the patterns computers find in who we’re attracted to. When I asked Justin Long, founder of the AI dating company Bernie.ai, what patterns his software found, he wouldn’t tell me: “Regarding what we learned, we had some disturbing results that I do not want to share. They were quite offensive.” I’d guess the findings were racist: OkCupid statistics show that even though people say they don’t care about race when choosing a partner, they usually act as if they do.

“I personally have thought about whether my swiping behavior or the people I match with reveal implicit biases that I’m not even aware that I have,” said Camille Cobb, who researches dating tech and privacy at the University of Washington. “We just use these apps to find people we’re interested in, without thinking. I don’t think the apps are necessarily leaking this in a way that would damage my reputation—they’re probably using it to make better matches—but if I wish I didn’t have those biases, then maybe I don’t want them to use that.”

Data thirst

Even if dating companies aren’t using our data to damage our reputations, they might be using it to make money. “It’s sketchy to think what type of information they could give advertisers, especially if it’s information we don’t even know about ourselves… I don’t smoke but maybe if I swipe right on a lot of guys who like cigarettes in my pictures, it reveals I think cigarettes make you look cool.” An advertiser could learn what products we find subconsciously sexy—literally—and show us targeted ads.

Yet these types of tailored recommendation algorithms all seek to make us right-swipe more. As apps truly get better at learning who we like and who we are, they may render swiping, liking, and messaging obsolete. This was the thought Canadian engineer Justin Long had when he built a “personal matchmaker assistant” called Bernie.ai. Frustrated by how much time he spent swiping and messaging compared to going on actual dates, he decided to build a bot to do the work for him. His app, Bernie, asked users to link their existing Tinder accounts and then watched them swipe, meanwhile modeling users’ individual tastes. Then Bernie started swiping on Tinder for them. If the AI encountered a mutual match, it would start a conversation with the opening line, “Do you like avocados?”

Tinder eventually forced Long to cease operation, but Long thinks personal dating assistants like Bernie are the future of dating tech. Instead of spending time swiping and messaging, we’ll give our digital matchmakers access to our calendars and GPS locations and let them deal with logistics on our behalves. Then, “my Bernie will talk to your Bernie,” says Long, and organize dates automatically. When algorithms are so good that we trust their decisions, perhaps we won’t mind giving them more control of our love lives. 

You’re on your own

As algorithms get better, they’ll need to collect data not just on whose profile photos we like but also who we feel chemistry with in person. Not a single dating app (that I’m aware of) asks users for the outcomes of actual dates. When I asked OkCupid’s Director of Engineer Tom Jacques (my old boss) why, he cites bias: “It’s a tricky issue because there is a very steep drop-off in what information people will volunteer, and we can only keep track of interactions between members while they are using the site. At some point, they will take their connection to the real world, and very few people who go on a date (successful or not) will tell us.” Yet we volunteer more than enough information for apps to be able to deduce how our dates went. They could use our GPS coordinates to watch who we go on dates with, how long those dates last, and whether they lead to a second date. The dating app Once even let daters monitor their heart rates on dates through their Fitbits to tell how much they found their date arousing. (Though Rosalind Picard, an expert on reading emotion from biosensors from MIT, told Gizmodo that changes in heart rate are more likely to reflect body movements rather than small changes in emotion.)

Today, dating apps don’t (openly) mine our digital data as nearly much as they could. Maybe they think we’d find it too creepy, or maybe we wouldn’t like what they learned about it. But if data mining were the key to the end of the bad date, wouldn’t it be worth it?

I’m still on the fence, but as much as I like the idea of a hyper-intelligent, perceptive dating algorithm, I think I’ll delete my Loveflutter account.

Dale Markowitz is a software engineer and data scientist living in New York City.

A high-tech dress made of Swarovski crystals that takes fashion and digital craft to atmospheric levels.

View Signal Source

It was approximately one year ago, when angry tweeters alleged that Jack Dorsey et al., were purposefully censoring and "suppressing" certain content on Twitter, namely anything to do with the leaked DNC and John Podesta emails, as well as hashtags critical of Hillary Clinton while "shadow-banning" pro-Donald Trump content. We can now confirm that at least one part of the above was true, because during today’s Senate hearing, Twitter admitted it "buried", which is another word for censored, significant portions of tweets related to hacked emails from the Democratic National Committee and Clinton campaign chair John Podesta in the months heading into the 2016 presidential campaign.

As Daily Caller’s Peter Hasson reports, Twitter’s systems hid 48 percent of tweets using the #DNCLeak hashtag and 25 percent of tweets using #PodestaEmails, Twitter general counsel Sean Edgett said in his written testimony to the Senate Judiciary Committee on Tuesday.

Before the election, we also detected and took action on activity relating to hashtags that have since been reported as manifestations of efforts to interfere with the 2016 election. For example, our automated spam detection systems helped mitigate the impact of automated Tweets promoting the #PodestaEmails hashtag, which originated with Wikileaks’ publication of thousands of emails from the Clinton campaign chairman John Podesta’s Gmail account.

 

The core of the hashtag was propagated by Wikileaks, whose account sent out a series of 118 original Tweets containing variants on the hashtag #PodestaEmails referencing the daily installments of the emails released on the Wikileaks website. In the two months preceding the election, around 57,000 users posted approximately 426,000 unique Tweets containing variations of the #PodestaEmails hashtag.

 

Approximately one quarter (25%) of those Tweets received internal tags from our automation detection systems that hid them from searches.

 

As described in greater detail below, our systems detected and hid just under half (48%) of the Tweets relating to variants of another notable hashtag, #DNCLeak, which concerned the disclosure of leaked emails from the Democratic National Committee.

And yet, this glaring act of censorship was not aimed at the sources of the alleged propaganda, but the content: Just 2% of the tweets using the #DNCLeak hashtag came from “potentially Russian-linked accounts,” Edgett said.

He also explained that Twitter hid the tweets as “part of our general efforts at the time to fight automation and spam on our platform across all areas.”

…And Hillary still lost?

Just over a year ago, on the same day that Donald Trump’s "grab them by the pussy tape" was released, WikiLeaks dumped over 30,000 hacked Podesta emails, which were damaging to Democratic nominee Hillary Clinton, throughout the election. A prior Wikileak of DNC emails, which revealed party officials secretly aided Hillary Clinton during her primary battle against Vermont Sen. Bernie Sanders , eventually cost then-DNC Chairwoman Debbie Wasserman Shultz her job. The leaks also exposed supposedly "neutral" journalists as pro-Clinton partisans.

The U.S. intelligence community concluded that Russian operatives were behind the original hacking of both the DNC and Podesta emails, which were part of Russian influence operations meant to disrupt the American electoral system.

*  *  *
A question emerges: did Jack Dorsey, with his arbitrary decision to censor specific content damaging to Democrats, interfere with the election, and a funnier question: if Hillary lost with Twitter censoring anti-Hillary content, what would the outcome have been if Twitter actually respected the First Amendment?

View Signal Source

View Signal Source

The likable psychopath is a great tool in any horror writer’s belt. Freddy Krueger. Chucky. Hannibal Lecter. Norman Bates. Pinhead.

Each of these villains finds a way into our hearts. Despite their intention to kill, often brutally, they possess such raw charisma that we, the audience, are willing to watch them wreak merry havoc on innocent people who we somehow like less than the literal murderer in our midst. Every example cited above has spawned numerous sequel films, books, TV series, and comics.

Written and animated by Alan Warburton with the support of Tom Pounder and Wieden + Kennedy, Goodbye Uncanny Valley explores the current state of photoreal CGI, where it came from and where it may be going to.

There are few things more bizarre than taking an inanimate object, giving it teeth, and setting it out into the world on a bloody rampage. Attack of the Killer Tomatoes brought us tomatoes; Rubber gave us a killer car tire; Death Bed: The Bed That Eats gave us, well, a bed that eats human flesh. And now we have killer donuts to contend with.