[Day 78] Love in the time of Artificial Intelligence

A couple of weeks ago, a Boston-based startup called Knowmail approached me, asking me to write about Artificial Intelligence for them. Somebody wants to pay me to write about AI? Man, I would pay people to listen to me rant about AI. The company pretty much gives me free range to write about whatever I want. For the first post, I wrote about what I would want to see in the next generation of dating. I know that if this app was available, I would use it. The original article can be found here.

learn365project_loveAI.png

Imagine you are on the subway and your phone tells you: “At 2 o’clock is a girl you’d find very attractive — a 9, in fact, and she may consider you above average. According to your expressed interests, she has 90% chance of being the one.”

You turn towards 2 o’clock. At the same time, the girl’s phone notifies her with a similar message. She looks up. Your eyes meet. You both smile. Sparks flow like electricity through the sweat-dampened air. Your phone asks whether you would like to share your contact information with her, and you say yes. It’s a match! You have just met the love of your life on the subway with the help of artificial intelligence.

Never again would you have to wonder if one of those who pass by unnoticed every day could be your special someone. Your phone constantly scans the crowd for faces you can fall in love with. No door is left unopened, no stone unturned.

Humans have spent thousands of years, perhaps millions, struggling in search of true love. From the historical likes of Antony and Cleopatra, to the fictional romance of Romeo and Juliet, we have lived and died for love. But what is love anyway? Is there any substantial answer to this question besides the tongue-in-cheek “Baby don’t hurt me, don’t hurt me, no more”?

Asking the so-called experts only yields the knowing smirk: “When you find it, you just know.” Should something as important as finding the one be left to chance? If Elon Musk can send a capsule to the vast nothingness of space and bring it safely home, shouldn’t we mortals be able to decipher the earthly mystery of love? Fortunately, thanks to recent advances in computer vision and machine learning, we are not all that far off. All of the technologies needed for the Tinder-meets-Skynet scenario described above already exist. Let’s walk through this scenario step by step.

First, you need badass face recognition technology. When you’re single and ready to mingle, just activate Love Mode, either pre-built into your phone or through a free app you can download, and your camera will scan every face in your peripheral vision. If that face doesn’t match any person in the database with Love Mode on, it simply passes. That person is probably not available anyway. But if that person happens to also be looking for love, the app will analyze that person for their attractiveness and compatibility with you.

State-of-the-art face recognition technology makes this task a cakewalk, no harder than Facebook’s automatic tagging. In the old days, when you uploaded a photo on Facebook, you would tag your friends manually. Nowadays, the site does it for you, just like magic. In a paper published in 2014, Facebook announced that, using a nine-layer deep neural network with more than 120 million parameters, their model can identify “faces in the wild” with almost 97.35% accuracy, closely approaching human-level performance. And that was 2 years ago.

Analyzing someone’s attractiveness is not as hard as many would think. “Beauty is in the eye of the beholder” is a lie, same as “Sorry, I didn’t check my phone until just now”. Researchers around the world have shown that beauty is, in fact, perfectly quantifiable, and comes with a straightforward formula. In 2007, a group of computer scientists at Tel-Aviv University used machine techniques popular in the 90s and early 2000s—linear regression with support vector machines—to predict someone’s attractiveness through their facial photos. They achieved a correlation of 82% with human ratings.

Furthermore, for an introductory course in Machine Learning at Stanford, my friends and I built a simpler system to predict attractiveness. Using solely spatial features of the face, our primitive models achieved an accuracy of 74.2%. If we have enough data from you, such as your own ratings of others, we can build a more sophisticated system that can rate someone’s attractiveness the way you would rate them.

Oh, and sorry to break it to you, but no matter how un-shallow you think you are, you probably do have an inner rating system. How else can you explain that time you turned down a great friend, whom you’d have said yes to on the spot had they been a little more attractive? Or how about that time when you couldn’t take your eyes off a random stranger on a train? What is “love at first sight” if not instant attraction? We all know that attractive people get paid more at work and have more success in dating. Everyone constantly judges how everyone else looks, even if it’s just on a subconscious level.

So how does your phone know that someone has a 90% chance of being The One? People can be married for years and still not know if their spouse is the one or not. Humans are excellent at recognition, intuition, and creating rapport, but machines beat us effortlessly on tasks that require number-crunching and consistent rationality. Machines can find your match the same way Google knows what kinds of ads to show in your inbox, Amazon knows what products to recommend, and OKCupid knows what profiles will catch your eye.

The app can pull your Facebook timeline, your search history, your Snapchat stories, and more. With all that data, it’s trivial to find out what movies you have watched, what kind of music you listen to, whether you support Trump, and how bad your daddy issues are. It then analyzes your target’s public profiles to get the same information, and computes your compatibility using a model trained on information from thousands of other couples. Given enough data, anything is quantifiable.

This analytical approach, though seemingly surreal, is not unheard of. The Internet is flooded with articles about nerdy, new-age men and women using algorithms to find love. Wired ran a 3000-word feature about the supposed math genius Chris McKinlay who hacked OKCupid to find true love, and it became one of Wired’s most read articles in 2014. The same year, Wall Street Journal reporter Amy Webb detailed the story of how she crunched data from various dating websites to find her eventual husband in the aptly titled book “Data, A Love Story: How I Gamed Online Dating To Meet My Match.” The book became wildly popular among women hoping to catch their big fish on eHarmony, and earned her a TED Talk. Both Amy and Chris did their work manually and religiously. Now, with artificial intelligence, your phone can do the dirty work for you, much faster, much more exhaustively, and with far better prospects. The result will be a date set up by an entity that knows you better than anyone ever has: your smartphone.

The biggest challenge for this scenario is not its feasibility, but the ethical questions it inevitably raises. Just because we can do something, should we? To start with, face recognition is extremely creepy. A camera can catch a glimpse of your face and tell exactly who you are, the places you frequent, whom you hang out with, and even when you had your last period (if you have them).

Would you be comfortable with that much exposure, in return for a significantly higher chance of finding someone whom you love and who will love you?

We live in a world where we have to constantly make the tradeoff between privacy and convenience. Should I write that email, knowing its content will be analyzed to optimize ads? Should I authorize that app, knowing it will know everyone I friend on Facebook? We already compromise much of our privacy every time we geotag an update on Facebook, make a purchase on Amazon, or enter a search query on Google. Since our privacy is already more of a mirage than a right, we might as well get some love out of it.

The next question is: does turning love into an equation take all the romance out of it? If this service existed, would you use it? All you’d have to do is flip a switch and let the machine take it from there. No need to sweat over making a date-worthy online profile. No need to answer hundreds of questions. No need to consciously swipe left or right on someone based on how they look. Or, would you rather wait for fate to bring that special someone to your workplace, or to that house party you didn’t want to go to but ended up attending anyway? It’s a decision each of us would have to make for ourselves. Looking at the rise of Tinder, OKCupid, and a slew of other dating apps, it seems like a lot of us have made up our minds already. One thing is certain: meeting that special someone on the subway, with the aid of your omnipotent AI assistant, is a much better love story than swiping right.

 

Advertisements
[Day 78] Love in the time of Artificial Intelligence

One thought on “[Day 78] Love in the time of Artificial Intelligence

  1. I agree with all the questions you posed on this topic. But imagine that we live in the future, 50 years from the advent of this dating application. I’d imagine the kind of questions we’d ask ourselves are:

    “How could I have ever met my boyfriend if it weren’t for X AI?”
    “How terrible would my life have been without X AI’s help on determining whether A or B was better for me? X AI was totally correct in predicting that B would turn out to be a disaster.”

    It could go either way. And I’d imagine that with people instinctively leaning on convenience more than privacy, perhaps the future isn’t so far from your predictions.

    Like

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s