tech

Japan Has a New Social Network But It's Hardly Letting Anyone In

If you can get past the AI bouncer, you're in.

by David Whelan
04 September 2015, 11:33am

In July of this year, a new Japanese social networking platform called Lemon was launched with the mission statement to connect like-minded people from universities all over Japan. So far, so Facebook. But Lemon differentiates itself in two key factors – one, it's only letting the top one percent of applicants in and, two, it has its own artificial intelligence system to stop the unworthy from signing up.

If you're not good enough, you're not getting in. And there's no way to sweet talk someone you know to get an invite or sleep with the creator. Social standing, potential, desirability and likeness to the existing member base are all factors taken into account by the AI in order to deem whether or not you are worthy.

The AI is all controlling and it knows whether you've been naughty or nice: it goes through your pre-existing social networking profiles, as well as your internet footprint. If you match up with at least 30 percent of Lemon's current users, you gain entry. It's not easy, believe me; I tried to sign up in the process of researching this piece and I got turned away. I felt like Patrick Bateman trying to get a table at Dorsia.

Its founder, Yusuke Matsumura, is what most people would call a whiz-kid. He completed a PhD in AI in a year and a half, and worked as a researcher for IBM's TJ Watson Research Center. In a recent interview with Tech Crunch Yusuke was quoted as saying "people cannot verbalize or understand their own preferences", which goes some way to explain what he believes the AI does. It helps people find what they didn't know they wanted.

But what interested me most about Lemon was its contrary nature: a social network without many people seemed counter-intuitive. I was also intrigued by the exact nature of the AI – whether, for example, it could be used for other, more nefarious means – so I reached out to Yusuke to get some answers.

VICE: Hi Yusuke. Can you tell me a little bit about how this started?
Yusuke: In November 2014, I founded Lip, alongside my co-founder and three engineers. We wanted to focus on social and matching services, and in January 2015 we received an investment of 40,000,000 yen [about £21m] from four celebrated Japanese VCs and an angel investor. Since then, after trying to make some prototypes, we hit upon the idea of Lemon.

Tell me about how that came about.
In our opinion, for social network services and matching services to succeed, the initial user base is of the most importance. So we wanted to create a space for users who are less interested in the smaller, refined functionality than the question of who is actually using the service.

So, this is where the AI comes in?
Our biggest trouble was in deciding on the method of inspection. Rather than a community of elites with amazing profiles, we wanted to provide a comfortable community for users.

What does "comfortable community" mean?
Instead of creating a fixed set of criterion for who can join, we made it so that it was decided based on the fixed number of existing users and whether there was affinity or not.


WATCH: A clinic in Los Angeles is using virtual reality to help heal war veterans


You create a space of like-minded people, then.
Lemon is a community that users [can] trust. We define Lemon as an online community for networking that allows people to link with new people without any anxiety. So, our ideal user is people who want to expand their network.

Can you tell me a little bit more about how it all works under the hood?
Our affinity identifying algorithm began with a combination of natural language processing and collaborative filtering.

That is, to the layman, an AI that is capable of deriving meaning from what humans do, and then using that to figure out who has things in common with Lemon users?
Yes, and while it is still now undergoing constant improvement, we are currently researching a way to make it work regardless of language to avoid problems when we expand Lemon overseas.

So, my rejection may have just been because of language. Phew.
Once we have finished our research on an algorithm that doesn't depend on language, we will start rolling out throughout Asia.

Our AI is not simply used for the admissions process, but also for deciding on a "recommended user" – which introduces user to user every day. In order for this to work, we adjust the algorithm daily. Then, this date is applied to the admission algorithm.

On MOTHERBOARD: The Two People at the Heart of Bitcoin's Current Crisis

Eventually meaning you discover the perfect Lemon user, over time. But could the AI work for other things? For example, could it be used to decide who gets into a club or who is the best candidate for a job?
Of course, I think it has lots of potential applications. On the business side, we've already received HR enquiries from companies based on the value of our users. In any rate, instead of adding functionality at random, we plan to carefully and deliberately work on finding our user's hidden needs.

Hidden needs?
For example, a lot of our users organise parties with each other, but I think it would be interesting if parties were planned by Lemon and members automatically invited.

What has the reaction to Lemon been like in Japan so far?
The reaction was stronger than we thought it would be.

Strong in what way? Don't you worry that if everyone is the same and like-minded, that the opportunity to debate and grow is lessened?
The exclusivity of Lemon has caused a lot of debate on sites like Twitter, and the truth is there are some who don't think it's a good thing. However, the admission process as it is, is only used at the very first step of forming a community, so keep in mind that our final objective is not just screening people. We just want to create our community healthily.

Thanks for chatting with me, Yusuke.

@MrDavidWhelan

More like this on VICE:

I Spent Two Years Trying to Be a Model on a Social Network

The Future of Social Media According to VICE

Who the Hell is Still Using Ello?