Shulou(Shulou.com)06/02 Report--

With the development of AI technology, its appearance in any field is not enough to attract the attention of the shock Department.

For example, the fourth season of "Black Mirror" depicts such a plot: in the future world, users can interact with the "AI dating system", the system will compare all kinds of data information, and finally find the perfect match rate of 99.8% through the algorithm.

AI pairing sounds simple, efficient and heartbeat, and this technology is not far away.

Foreign Viola.AI, Coffee Meet Bagel, Bernie, Loveflutter, domestic Jiayuan, Baihe.com, Momo, NetEase Huatian, etc., all have their own artificial intelligence matching systems, which can match according to the age, address, annual salary, interest, consumption habits and other big data entered by the user.

However, in real application, AI matching also has a lot of speechless places.

Find true love by AI: it looks beautiful

At present, there are two kinds of matching methods for the mainstream AI dating platform on the market: one is to improve the matching level by fully mining and predicting the data submitted by social networks and users. One is to narrow the gap between online and offline attraction by combining biological means such as facial recognition and DNA detection.

The combination of the two is more common to decipher the love password. For example, Pheramor, a company whose core business is to facilitate matches through users' DNA, also trains artificial intelligence to mine users' social media to understand the personalities of both sides.

But the reality is not as perfect, simple and easy to go up as depicted on the corporate website.

First of all, there is some distortion in social media data. The information provided by individuals or presented on the network will inevitably be different from the real personality. AI can not make a comprehensive and in-depth user portrait in semantic understanding, and the accuracy of matching naturally can not stand scrutiny.

What's more, online dating platforms have no incentive to create and optimize perfect matching algorithms. Because if the algorithm matching degree is very high, users will often "run" after falling in love with each other, and the platform will lose its profit space. So the best way for business is to allow a certain deviation rate to introduce objects that may be good enough but not perfect, and after several unsatisfactory dates, the user will go back to the platform to start the next attempt.

And what if there is a dating platform with boundless love, with the belief that every pair of predestined cp will be put into marriage? From online matching, to making connections, getting to know each other, to offline dating, and finally into marriage, this is obviously a process with a very long chain and full of uncertainty. Only in the matching stage, it is difficult to say that AI can bring any subversive help to the quality of human love and marriage. Bernie.AI stopped the project after about 100000 screening pairs.

Is there a more useful AI tool that can delve into more critical aspects of intimacy?

Fall in love at first sight, you may need an AI wingman.

Going out on a blind date is not a pleasant experience, so that contemporary "social phobia" young people are not very active in finding a partner. The Japanese government conducted a questionnaire survey of 3980 men and women who wanted to get married, and the results showed that more than half of them did not take any action. Young single men over the age of 20 are waiting for their partner to show up.

In this context, there is even a robot dating business, which inputs the answers to all questions in advance, and the self-introduction, hello and chat are done by the robot, so that you don't have to worry about saying the wrong words, and you can easily get each other's messages. It's just a little weird in the picture.

This robot is a bit like the secret weapon often used by love experts-wingman.

A wingman originally refers to a plane that stays in a designated position in a flight formation and carries out orders from the captain, but in the course of dating, what the wingman often has to do is to score points for the protagonist.

For example, if you are worried about leaving a bad impression, let the "wingman" exchange first, paving the way for further contact with people who want to chat up. Once there is a cold spot, the wingman will be the one to break the deadlock. If the protagonist has a high conversation, "wingman" will naturally become a quiet beautiful girl / man.

Having a reliable "wingman" can indeed resolve a lot of embarrassment, and at a critical time, it can also help two strangers get to know each other better, which is obviously much more effective than a simple matching of conditions.

But human emotional communication is really complex and unpredictable, and wingmen need the ability to "understand the air" (meaning sensitive to the atmosphere). Can AI shoulder this important task?

At the very least, many people are already trying to reach it.

Last summer, dating platform Match partnered with Google to create a dating robot called Lara. It extracts the profile of a potential partner every day, and if the system detects a common interest, it provides further advice on where to have a drink, what someone's most important first social document is, and how to plan the next day's itinerary.

Loveflutter, a dating platform from the UK, also uses artificial intelligence to guide users on how to communicate more comfortably during offline meetings after analyzing users' tweets and chat records.

AIMM, a start-up owned by ICA, is guided by dynamic communication between users. When a user expresses interest in a potential partner, AIMM will call you first and ask artificial intelligence to guide you on what to say, sometimes to keep you calm, and sometimes to tell you who the other person is and what lifestyle he or she likes. If someone often mentions money, the system will infer that financial ability is more important to them.

At the end of the conversation, AI will provide feedback based on the overall feeling. If users think highly of this communication, the system will follow up and set a new date, gently motivating you to "move on." If the other person likes "tradition", AI may advise you to go for a walk in the park, while for a "trendy", rock climbing may be more suitable.

In terms of encouragement, AI of the AIMM family is also very kind and will not directly say negative words like "girls don't like you" to men. When arranging the first date, you will tell the user, "Don't worry, I will be with you every step of the way, and I will try to make it easy to attract the other person."

("Black Mirror" stills: men and women make their first date through the love system.)

Does it sound a bit like Coach, the voice assistant in Black Mirror that helps the protagonist review and simulate the emotional route?

Andrew McStay, a professor of digital media at the University of Bangor in the UK, agrees that artificial intelligence is the future of dating, while voice-based dating programs may be more "natural". In the words of Kevin Terman, founder of AIMM, "the traditional online dating industry has gone bankrupt."

What are the problems with the emergence of AI "wingman"

Of course, you might ask, is it true that AI won't screw up a date that you finally got together? A reliable "AI wingman" naturally requires it to achieve a certain accuracy in some basic abilities, otherwise it will not do too much harm to single animals if they misunderstand each other's meaning.

So, what kind of technical training is needed for this God's assists to perform perfectly?

1. Emotion recognition of Excelsior

"AI wingman" is not so much a dating problem as a human-computer interaction problem. With the improvement of natural language processing technology, especially the application of emotion analysis in multimodal interaction, it can give effective guidance subjectively. For example, a video component has been added to AIMM, which can mobilize the camera to analyze the facial reactions of both sides, and combine voice to determine the level of interest in the next move. If a person says "I am interested" because of a tired day at work, it is likely to be misjudged by the system as a lack of interest, and the addition of facial expression information can effectively supplement the relevant details of emotional inference.

On the whole, machines are not good at explaining emotions according to people's expressions, and some emotional nuances need to be judged comprehensively by human input, background and history, which is difficult to do for the time being.

two。 Algorithm bias that is easy to be ignored

At present, the mainstream intelligent dating software is mostly designed for a man to ask a woman out, so the algorithm can easily be tuned by the online text big data into a potential discriminator. AIMM's artificial intelligence tends to provide only male information and advice for first dates. Other designs are also unfriendly to LGBTQ groups, who are more likely to use online dating to find partners. Eliminating these biases technically can undoubtedly help online dating programs gain an edge in the competition.

3. Terminal computing that is difficult to be dual-complete

In the process of AI-assisted love relationship, the terminal hardware carries many calculations needed for real-time identification, judgment, decision-making and so on. In other words, in order to solve complex emotional interaction problems, the demand for computing power is increasing as never before. On the terminal, computing resources and power consumption are very limited. For example, many dating AI are supported by smart devices in the home scene. For example, Match's love assistant Lara is activated by Google Home. There are two key issues involved. One is the computing cost. Uploading data to the cloud computing can greatly reduce the computing cost and avoid the high investment of smart chips on the speakers; the other is the risk. Dating, love and other issues involve a lot of personal privacy. Cloud computing how to ensure the security of user information naturally becomes more difficult for the platform. Who can organically integrate the two, so that the terminal can safely and efficiently support the dating guidance services that users need at any time, will be able to seize the market opportunity.

Generally speaking, in the sinking stage of AI, the emergence of new technologies often need to be discussed, not the right and wrong topics such as "whether AI is possible or not", but the argumentation of "how many points can AI do?"

If you want to build a bridge to love with AI, you may not need creativity, but step by step to explore the boundaries of technology, cost, production path, and other engineering details. A small step in human dating is a big step in AI.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.