Connect with us


Support the Guardian and help us highlight the rise of digital inequality



Automating Poverty, our year-long project about the takeover of welfare systems by algorithms and AI, was funded with reader donations raised last year

In countries across the world, algorithms and artificial intelligence are taking over welfare payments systems. Governments often make these changes quietly, with little public debate or accountability. But they affect millions, with serious and possibly even fatal results.

In our special project Automating Poverty, we cast a light on the way digital innovation is threatening the poor.

Our correspondent Rebecca Ratcliffe travelled 1,300km from Delhi to the east of India to investigate the countrys vast biometrics scheme, Aadhaar. She interviewed the family of Motka Manjhi, who died from starvation after his food subsidies were stopped because his thumbprint wasnt recognised by the Aadhaar biometrics database.

In Australia, Luke Henriques-Gomes reported on the families who are informed by text message that their payments are suspended, sometimes in error and with no human to complain to.

In the UK, Robert Booth and Sarah Marsh spent the best part of six weeks digging into the automation of welfare systems.

After publishing our series, we received scores of emails from social scientists, government officials and welfare recipients around the world. Writers from India, Spain, the UK, the US and other countries shared their experiences.

At the United Nations general assembly in New York, the Guardian mediated a panel discussion on the human rights challenges of the digital age.

Automating Poverty is one of the reporting projects we funded with the $1m in reader donations we raised during our end-of-year drive last winter.

Now through January, we hope to raise $1.5m to fund more journalism like this in 2020. With your help, we will continue to fight for the progressive values we hold dear democracy, civility, truth.

Please consider making a contribution. And as always, thanks for reading.

Read more:


‘This is small talk purgatory’: what Tinder taught me about love




When I ended up single in a small town, I turned to a dating app. But finding someone fully and messily human was harder than I thought

I did not intend to be single in the rural village where I live. Id moved there with my fiance after taking a good job at the local university. Wed bought a house with room enough for children. Then the wedding was off and I found myself single in a town where the non-student population is 1,236 people. I briefly considered flirting with the cute local bartender, the cute local mailman then realised the foolishness of limiting my ability to do things such as get mail or get drunk in a town with only 1,235 other adults. For the first time in my life, I decided to date online.

The thing about talking to people on Tinder is that it is boring. I am an obnoxious kind of conversation snob and have a pathologically low threshold for small talk. I love people who fall into the category of Smart Sad People Flaunting Their Intelligence With Panache. I love Shakespeares fools and Elizabeth Bennet and Cyrano de Bergerac. I love Gilmore Girls and the West Wing and Rick And Morty. I want a conversation partner who travels through an abundance of interesting material at breakneck speed, shouting over their shoulder at me: Keep up. I want a conversation partner who assumes I am up for the challenge, who assumes the best of me.

It will not surprise you to learn that this is a totally batshit way to approach Tinder and that, for my snobbery, I paid a price.

The first man I chatted with who met my conversational standards was an academic, a musician. He taught refugee children how to play steel drums. He had a dark sense of humour, he was witty, and he laid all his baggage out there on the line right away. Even through our little chat window it was obvious he was fully and messily human, which I loved, and so we chatted all day long, for days, and I could not wait to meet him.

Reality was different. What had seemed passionate and daring online, turned out to be alarmingly intense. There were multiple bouts of tears, there were proposed road trips to Florida to meet his mother and dog, there was an unexpected accordion serenade, and there was the assertion that I would make a very beautiful pregnant woman. Listen: I think a man who can cry is an evolved man. I hope to some day have kids, which, I suppose, would entail being, for a time, a pregnant woman. I even like the accordion. None of this was bad on its own, but it was so much. After I said I didnt want to date any more he sent me adorable letterpress cards in the mail with upsetting notes inside that said he was upset, no, angry, that I wouldnt give us a shot.

I chalked this experience up to bad luck, and continued to only date people with whom I had interesting online conversations.

My next IRL date had just moved to New York by way of Europe and was a collector of small stories and observations. Our chats took the form of long blocks of text. Anecdotes swapped and interrogated. Stories from the world presented to each other like offerings dropped at each others feet. I love such things; I am a magpie at heart.

But these stories became grotesque in real life. My date spent most of our dinner conversation monologuing about how Americans were very fat, which made it difficult to enjoy my chiles rellenos. But when we went back to his apartment for a drink, it was beautifully decorated: full of plants and woven hangings and a bicycle propped against a shelf full of novels. He was smart and handsome and sort of an asshole, but perhaps in a way that would mellow over time in a Darcy-ish manner. We drank some wine and eventually I said I should go home but he got up and kissed me, kissed me well, so I told myself this was what online dating was like, and I should carpe diem and have an experience.

During sex, he choked me. Not for long, and not very hard, but his hands manifested very suddenly around my throat in a way I know was meant to be sexy but which I found, from this relative stranger, totally frightening. I had not indicated this was something I liked, and neither had he. I know people are into that. I could even be into that. But not as a surprise.

Afterwards, he chatted to me as I counted the appropriate number of minutes I needed to wait before making an exit that wouldnt seem like I was running away. He said that he was really interested in mass shooters and the kinds of messages they left behind and, still naked in bed, he pulled out his phone and showed me a video from 4Chan. It was a compilation of mass shooters video manifestos, but set to comically upbeat music. Its hilarious, he asserted. I said I had to go. The next day, and a few times after, he messaged asking why I had run away and gone dark.

I realised that perhaps what seemed interesting online did not translate into real life. My method of going on dates only with people who gave good banter was working poorly. It was pointing me toward the extremes.

But once I gave up on the banterers, my Tinder chats became uniform. The conversations read like a liturgy: where are you from, how do you like our weather, how old is your dog, what are your hobbies, what is your job, oh no an English teacher better watch my grammar winkyfacetongueoutfacenerdyglassesface. The conversations all seemed the same to me: pro forma, predictable, even robotic.

Thats when I realised that what I was doing amounted to a kind of Turing test.

This seems a good moment to tell you that, for a civilian, I know a lot about robots. Specifically, I know a lot about chatbots and other AI meant to perform their humanity through language. In fact, I was teaching undergrads about robots in science writing and science fiction when I began online dating. In class, we discussed the ways in which a robot, or chatbot, might try to convince you of its humanity. This effort is, in short, called a Turing test; an artificial intelligence that manages, over text, to convince a person that it is actually human can be said to have passed the Turing test.

I began seeing similarities between the Turing test and what us Tinder-searchers were doing whether we were looking for sex or looking for love. A Tinder chat was its own kind of test one in which we tried to prove to one another that we were real, that we were human, fuckable, or possibly more than that: dateable.

Online dating seemed more bearable when I thought of it this way. It was easier to pretend I was a woman conducting a scientific investigation of language and love than it was to admit I was lonely. Easier than admitting that an algorithm someone had made to sell ads to singles was now in charge of my happiness. Easier than admitting that this was a risk I was willing to take.

I knew a little bit about how to proceed with my Tinder Turing tests from one of my favourite books one I was teaching at the time: The Most Human Human, by Brian Christian. In this book, which I have read five times, Christian goes to participate in the worlds most famous Turing test, the Loebner prize in Brighton. He serves as a human blind, chatting with people through an interface, who then have to decide whether he is a human or a chatbot. The true point of the Loebner prize is to see whether any of the chatbots can convince the judges of their humanity but as Christians title suggests, there is also a jokey prize offered to the human blind who the fewest participants mistake for a robot. Receiving the Most Human Human award was Christians goal. In the book, he asks: what could a human do with language that a robot could not? What are the ways of expressing ourselves which are the most surprisingly human? How do we recognise our fellow humans on the other side of the line? And so, as I attempted to find the lovely and interesting people I was sure were lurking behind the platitudes the average Tinder chat entails, I asked myself Christians question: how could I both be a person who understood she was online, on Tinder, but still communicate like a humane human being? What could I do that a robot couldnt?

I was thinking of robots metaphorically, but there are real chatbots on Tinder. I never encountered one (to my knowledge; was Dale, age 30, with the six pack and swoopy hair and the photo on a yacht who wanted to know if I was DTF RN only ever just a beautiful amalgamation of 1s and 0s?). But I know lots of people who have, and men seem to be particularly besieged by them. This is such a common problem on Tinder that a culty test has emerged a kind of CAPTCHA for humans to deploy if a match seems suspiciously glamorous or otherwise unreal. In the Potato test, you ask the person youre speaking to to say potato if theyre human. And if they dont, well, you know. You might think this is ridiculous but one of my favourite screen shots of this going down (the Tinder subreddit is a glorious place) reads as follows:

Tinder: You matched with Elizabeth.
Actual Human Man: Oh lord. Gotta do the Potato test. Say potato if youre real.
Elizabeth: Heyy! you are my first match.
I dare you to try to make a better first message ahaha.
Actual Human Man: Say potato Elizabeth.
Elizabeth: And btw, if you dont mind me asking this, why are you on Tinder?
Personally I think Im not much into serious stuff ahaha.
Actual Human Man: SAY POTATO.

Meanwhile, the conversations I was having with true potato-tested men and women werent much different from Actual Human Mans conversation with Elizabeth. These conversations never resolved into anything more than small talk which is to say they never resolved into anything that gave me a sense of who the hell I was talking to.

I started taking hopeful chances again, and many of my conversations yielded real-life dates. I could write you a taxonomy of all the different kinds of bad those dates were. Sometimes it was my fault (blazing into oversharing and rightfully alienating people), sometimes it was their fault (bringing his own chicken sandwich and commenting on my tits within the first 15 minutes), and sometimes it was nobodys fault and we had a fine time but just sat there like two non-reactive elements in a beaker. One way or another, though, what it always came down to was the conversation.

The chapter I have always loved most in Christians book is the one about Garry Kasparov losing at chess to Deep Blue, IBMs chess-playing computer. Christian explains the chess concept of playing in book. In short, the book is the known series of chess moves that should be played in sequence to optimise success. In most high-level chess matches, the first part of any game is played in book and a smart observer will know which moves will follow which until a certain amount of complexity and chaos necessitates improvisation at which point the players begin to play in earnest. Some might say, as themselves. Kasparov holds that he did not lose to Deep Blue because the game was still in book when he made his fatal error and so, while he flubbed the script, he never truly even played against the algorithmic mind of his opponent.

In this chapter, Christian makes a brilliant comparison between most polite conversation, small talk, and the book, arguing that true human interaction doesnt start happening until one or both of the participants diverge from their scripts of culturally defined pleasantries. The book is necessary in some ways, as it is in chess (Bobby Fischer would disagree), in order to launch us into these deeper, realer conversations. But it is all too easy to have an entire conversation without leaving the book these days to talk without accessing the other persons specific humanity.

This was my trouble with Tinder. No matter how hard I tried to push into real human terrain over chat, and sometimes on real-life dates, I always found myself dragged back into a scripted dance of niceties. I might as well have been on dates with Deep Blue, ordering another round of cocktails and hoping its real programming would eventually come online.

After these dates, I felt pretty low. Like I would never find what I was looking for.

What was I looking for?

To answer that, I have to go back to Elizabeth Who Wouldnt Say Potato. Theres something about the way her suitor asks her not if shes human, but if shes real, that Im a sucker for. Theres a passage from The Velveteen Rabbit that my sister asked me to read at her wedding. I thought I was up for the task (its a childrens book, for Gods sake), but when the time came, I ugly-cried all the way through:

Real isnt how you are made, said the Skin Horse. Its a thing that happens to you. When a child loves you for a long, long time, not just to play with, but REALLY loves you, then you become Real.

Does it hurt? asked the Rabbit.

Sometimes, said the Skin Horse, for he was always truthful. When you are Real you dont mind being hurt... You become. It takes a long time. Thats why it doesnt happen often to people who break easily, or have sharp edges, or who have to be carefully kept. Generally, by the time you are Real, most of your hair has been loved off, and your eyes drop out and you get loose in the joints and very shabby. But these things dont matter at all, because once you are Real you cant be ugly, except to people who dont understand.

Margery Williams Bianco, The Velveteen Rabbit

I want to pretend that Im cooler than crying about The Velveteen Rabbit but Im just not. And if Im honest with myself, this was what I wanted: for someone not only to prove to me that they werent a robot, but that they were real, and would make me real, too. Could I put this in my Tinder bio? CJH, 34: looking to keep it real and love off most of your hair till your eyes drop out <3.

It had been, by this point, a year of on and off Tinder dating. At one point I even googled Christian to see if he was single. He was not. On what I decided had to be my last Tinder date ever, a neuroscientist in a hipster diner delivered a nonstop monologue about his recent life that was mostly his consideration of moving to LA because the women there were so hot. He gave me a briefing on the various types of plastic surgery that were in right now. It was a conversation that felt like the headlines of checkout aisle magazines had come to life, to shame me for my non-cyborg womanhood.

Thats it, I told my friends, for whom I always performed the stories of my bad dates. Im done. Im ghosting everyone in my inbox and deleting my account.

I meant to.

But there was one man who kept talking to me.

Me: Im laughing at the part of your bio where you say youre hopelessly extroverted. Are you the sort of person who makes friends on airplanes?

Him: No but Im a chronic oversharer!

Me: Ive actually grown into oversharing. Its the only way to avoid infinite small talk purgatory.

Him: Tinder is by definition small talk purgatory.

Me: God save us all.

Him: Were all doomed.

Me: How do we escape?

Him: Get away from cell signals and head for the hills.

We were out of book. It was as if he had gestured to the conversational matrix we were talking inside of, the one Id been trying to escape, and said: hey, I see it, too.

Every day we kept talking and every day I said I was going to delete the app, but didnt. Because every time I tried, I wound up having delightful conversations with this human on the other side of the wires and waves. We developed our own language. There were inside jokes, callbacks, patterns of engagement. After that first day, a robot could not have replaced either of us, because our speech was for each other. It revealed who we were together: goofy, honest, heartbroken, funny about our sadness, a little awkward. The language we spoke in was what Christian would call site specific, meaning it was a language meant to exist in a certain place, at a certain time, with a certain person. It was the opposite of everything No Potato Elizabeth had to say.

Eventually, I agreed to go on a real-life date bargaining us down from dinner to drinks because my expectations were so warped and strange by this point. I made no effort to look nice. I drank two beers with friends beforehand to numb myself to the misery I anticipated. But as soon as I showed up at the brewery wed picked, I immediately regretted these decisions. The man sitting across the bar was even cuter than Id anticipated and, as I approached him, thinking about our conversations over the past weeks, I was able to admit to myself how much I hoped he might like me. How much I hoped I hadnt already blown this. As soon as we started talking, my ratty shirt and snowboots, my buzz and other defences, didnt matter, though. Our date was all of the things our chats were awkward, funny, honest, and backandforthy, which is to say: human.

I actually hate this brewery, I told him. Their beer is so bad.

Me, too! he said.

Then why did we pick it!

It just seems like the sort of place youre supposed to meet.

This past year, on our first anniversary, this man gave me a present. It was a blanket, and woven into it was the image of our first Tinder conversation. He laughed very hard, and I laughed very hard, as he offered it to me, because it was ridiculous. It was meant to be. But it was undercover earnest, too. It was sweet and it was dumb and I could not have loved that blanket more.

We split up before we could reach another anniversary, but as I went about the breakup torture that is boxing up all your exs things, the photos and gifts too painful to stare down, I couldnt give up the blanket. It was a reminder that being human is risky, and painful, and worth doing. That Id rather lose everything as Kasparov than succeed as Deep Blue.

The conversation on the blanket is actually quite long. You cant read precisely what it says, but you can see the rhythm of it. The longer bursts of sharing. The questioning responses. The patter. One of our friends, upon seeing the blanket, teased us. You talked for this long before you locked it up? You both need better game.

Its true that neither of us had any game. Its also true that this wasnt the point. The point was that we found a mutual language in which to prove ourselves human and pass each others Turing tests. We both understood how easy it is to let your life pass along, totally in book, unless you take a risk, and disrupt the expected patterns, and try to make something human happen.

If you would like a comment on this piece to be considered for inclusion on Weekend magazines letters page in print, please email, including your name and address (not for publication).

Read more:

Continue Reading


The AI stack thats changing retail personalization




Consumer expectations are higher than ever as a new generation of shoppers look to shop for experiences rather than commodities. They expect instant and highly-tailored (pun intended?) customer service and recommendations across any retail channel.

To be forward-looking, brands and retailers are turning to startups in image recognition and machine learning to know, at a very deep level, what each consumer’s current context and personal preferences are and how they evolve. But while brands and retailers are sitting on enormous amounts of data, only a handful are actually leveraging it to its full potential.

To provide hyper-personalization in real time, a brand needs a deep understanding of its products and customer data. Imagine a case where a shopper is browsing the website for an edgy dress and the brand can recognize the shopper’s context and preference in other features like style, fit, occasion, color etc., then use this information implicitly while fetching similar dresses for the user.

Another situation is where the shopper searches for clothes inspired by their favorite fashion bloggers or Instagram influencers using images in place of text search. This would shorten product discovery time and help the brand build a hyper-personalized experience which the customer then rewards with loyalty.

With the sheer amount of products being sold online, shoppers primarily discover products through category or search-based navigation. However, inconsistencies in product metadata created by vendors or merchandisers lead to poor recall of products and broken search experiences. This is where image recognition and machine learning can deeply analyze enormous data sets and a vast assortment of visual features that exist in a product to automatically extract labels from the product images and improve the accuracy of search results.

Why is image recognition better than ever before?


While computer vision has been around for decades, it has recently become more powerful, thanks to the rise of deep neural networks. Traditional vision techniques laid the foundation for learning edges, corners, colors and objects from input images but it required human engineering of the features to be looked at in the images. Also, the traditional algorithms found it difficult to cope up with the changes in illumination, viewpoint, scale, image quality, etc.

Deep learning, on the other hand, takes in massive training data and more computation power and delivers the horsepower to extract features from unstructured data sets and learn without human intervention. Inspired by the biological structure of the human brain, deep learning uses neural networks to analyze patterns and find correlations in unstructured data such as images, audio, video and text. DNNs are at the heart of today’s AI resurgence as they allow more complex problems to be tackled and solved with higher accuracy and less cumbersome fine-tuning.

How much training data do you need?

Read more:

Continue Reading


Lawyers hate timekeeping Ping raises $13M to fix it with AI




Counting billable time in six-minute increments is the most annoying part of being a lawyer. It’s a distracting waste. It leads law firms to conservatively under-bill. And it leaves lawyers stuck manually filling out timesheets after a long day when they want to go home to their families.

Life is already short, as Ping CEO and co-founder Ryan Alshak knows too well. The former lawyer spent years caring for his mother as she battled a brain tumor before her passing. “One minute laughing with her was worth a million doing anything else,” he tells me. “I became obsessed with the idea that we spend too much of our lives on things we have no need to do — especially at work.”

That’s motivated him as he’s built his startup Ping, which uses artificial intelligence to automatically track lawyers’ work and fill out timesheets for them. There’s a massive opportunity to eliminate a core cause of burnout, lift law firm revenue by around 10% and give them fresh insights into labor allocation.

Ping co-founder and CEO Ryan Alshak (Image Credit: Margot Duane)

That’s why today Ping is announcing a $13.2 million Series A led by Upfront Ventures, along with BoxGroup, First Round, Initialized and Ulu Ventures. Adding to Ping’s quiet $3.7 million seed co-led by First Round and Initialized last year, the startup will spend the cash to scale up enterprise distribution and become the new timekeeping standard.

I was a corporate litigator at Manatt Phelps down in LA and joke that I was voted the world’s worst timekeeper,” Alshak tells me. “I could either get better at doing something I dreaded or I could try and build technology that did it for me.”

The promise of eliminating the hassle could make any lawyer who hears about Ping an advocate for the firm buying the startup’s software, like how Dropbox grew as workers demanded easier file sharing. “I’ve experienced first-hand the grind of filling out timesheets,” writes Initialized partner and former attorney Alda Leu Dennis. “Ping takes away the drudgery of manual timekeeping and gives lawyers back all those precious hours.”

Traditionally, lawyers have to keep track of their time by themselves down to the tenth of an hour — reviewing documents for the Johnson case, preparing a motion to dismiss for the Lee case, a client phone call for the Sriram case. There are timesheets built into legal software suites like MyCase, legal billing software like TimeSolv and one-off tools like Time Miner and iTimeKeep. They typically offer timers that lawyers can manually start and stop on different devices, with some providing tracking of scheduled appointments, call and text logging, and integration with billing systems.

Ping goes a big step further. It uses AI and machine learning to figure out whether an activity is billable, for which client, a description of the activity and its codification beyond just how long it lasted. Instead of merely filling in the minutes, it completes all the logs automatically, with entries like “Writing up a deposition – Jenkins Case – 18 minutes.” Then it presents the timesheet to the user for review before they send it to billing.

The big challenge now for Alshak and the team he’s assembled is to grow up. They need to go from cat-in-sunglasses logo Ping to mature wordmark Ping.  “We have to graduate from being a startup to being an enterprise software company,” the CEO tells meThat means learning to sell to C-suites and IT teams, rather than just build a solid product. In the relationship-driven world of law, that’s a very different skill set. Ping will have to convince clients it’s worth switching to not just for the time savings and revenue boost, but for deep data on how they could run a more efficient firm.

Along the way, Ping has to avoid any embarrassing data breaches or concerns about how its scanning technology could violate attorney-client privilege. If it can win this lucrative first business in legal, it could barge into the consulting and accounting verticals next to grow truly huge.

With eager customers, a massive market, a weak status quo and a driven founder, Ping just needs to avoid getting in over its heads with all its new cash. Spent well, the startup could leap ahead of the less tech-savvy competition.

Alshak seems determined to get it right. “We have an opportunity to build a company that gives people back their most valuable resource — time — to spend more time with their loved ones because they spent less time working,” he tells me. “My mom will live forever because she taught me the value of time. I am deeply motivated to build something that lasts . . . and do so in her name.”

Read more:

Continue Reading