Technological and social trends of 2018

Main technological events and processes

Big data and neural networks are evolving

This year, it became noticeable how big data technologies began to influence the economic situation, when working with which machine learning, neural networks and pattern recognition are used.

According to experts, the size of the big data market is increasing every year. It is already the most dynamically developing area in IT. According to IDC analysts, by 2020 this market will grow to $ 203 billion. At forums dedicated to big data, which took place in Russia in the past year, they talked about how to use technology in business, science and education. So if you are deciding which profession to master or what to invest in, machine learning is a likely choice.

Neural networks attract the most attention of Internet users when they master the next impressive visualization or perform specific tasks, analyzing large amounts of data. For example, one algorithm learned to paint a detail of a picture in the same style as its background, another - to remove this very background from portraits of people (no more magnetic lassos in Photoshop), and the third even diagnoses depression by speech.

2018 Tech and Social Trends: Background Removal from Portrait

Robots are becoming more common

Automation is already affecting the service industry we encounter on a daily basis. And the cheaper technology gets, the more likely it is that many workers will be replaced by artificial intelligence.

If until recently shops without sellers seemed like something futuristic, today robotic checkouts appear in many large supermarkets.

Some companies are planning to completely abandon the services of live cashiers. For example, Amazon has already opened seller-free stores in US office centers and will expand to airports next year. Cold calls, which are rarely liked by both call center operators and service customers, are increasingly being taken over by machines. Robots call to tell about debts, offer goods and services, invite them to work, and chat bots advise users on websites.

If in the outgoing year they were more often worried that robots would take away jobs from people, in the next year it will be more about creating new vacancies. However, to remain in demand in the labor market, it will not be enough to mechanically process information and perform simple tasks. We'll have to develop creativity and not forget about humanity.

For example, according to a recent study by Cornell University, it is too early to replace lawyers with robots precisely because law is based on values and compromises, and artificial intelligence has yet to grow and grow to such abstract categories.

New technologies promise a renewed interest in ethics. In 2018, there was a second fatal accident involving an unmanned vehicle. The first was in 2016 with a Tesla car. It is not for nothing that the minecart problem has become memetic since then - an ethical conundrum that suggests deciding whether to save the lives of several people by sacrificing one. Even if people disagree about this task, cars that have to make decisions in emergency situations, all the more, raise many questions.

2018 Technological and Social Trends: Accident with Self-Driving Car and Cyclist

Smartphones are replacing the desktop

According to analysts from the communications agency Media Direction Group, the future belongs to mobile technologies. About 92% of users around the world access the Internet from a smartphone, of which about 85% - every day. In Russia, this figure is 54%, with 16% using only mobile Internet.

A full-fledged Internet of Things is on the doorstep. For example, the brand of chips Tostitos has released a trial series of packs with a breathalyzer and an NFC chip that can call a taxi for the user if the level of alcohol in his blood rises (strange that this initiative did not occur to alcohol brands). And mobile applications allow you to control your home appliances. For example, with Redmond technologies it is possible to regulate the amount of electricity consumed.

Even television is following the path of interactivity and user regulation via smartphones. At the beginning of 2018, the detective series Mosaic was released on HBO, where the viewer can choose a POV character, that is, decide whose eyes to look at events. And in the last days of the year, a new season of "Black Mirror" from Netflix will be released, one of the episodes of which will contain an episode with an interactive ending. To take advantage of these features, stories must be managed using apps.

Here are some of the technological developments and developments in 2018 that seem particularly significant and promising.

1. Launch of Google speech synthesis, based on neural network models, which generates speech that is almost indistinguishable from human speech.

Examples of using: voice robots indistinguishable from humans; call center robots. Previously, the technology of realistic speech synthesis was available only to large companies that created their own neural network models for speech synthesis. But the emergence of a cloud service from Google opens up the possibility of using this technology and small organizations. There is no cloud solution for the Russian language yet, but you can try it out.

2. Google has opened source the library for "Reinforcement Learning".

Examples of using: self-learning bots in games, self-learning business processes in companies. An online store that learns itself to sell products in the most efficient way, formulating hypotheses about who, when and how to recommend a particular product.

3. A new recurrent neural network architecture from Google called and, in general, the development of architectures that are inheritors of the Transformer model.

Examples of using: more advanced machine translation, chatbots, speech recognition. The emergence of these architectures marks another significant step forward in solving the problems of machine translation and understanding the meaning of the text.

4. The first prototype of a headset for silent speech recognition. Arnav Kapoor, a researcher at the Fluid Interfaces group at the Massachusetts Institute of Technology's media laboratory, demonstrated a prototype of a headset system capable of recognizing words spoken by a person to himself using an electromagnetic field sensor and a convolutional neural network.

Examples of using: the ability to mentally give the smartphone a command to perform calculations, turn on a new music track, tell the current time, exchange rates, read the latest news, and all this - without attracting attention to yourself and without breaking the silence. By combining such a device with augmented reality glasses, you can create a system that will, on a mental command, remember or find faces in memory, help navigate, give advice, and so on.

5. Powered by NVidia: Pretraining neural networks for real-world image recognition using auto-generated renders.

Examples of using: training robots (including car autopilots) on much smaller datasets. In essence, the robot learns using a specially generated realistic virtual space.

6. The emergence of many tools based on deep convolutional neural networks and generative adversarial networks (GAN).

Examples of using: just imagine a Photoshop-like application from the future that can automatically split an image into layers, give them names with a certain meaning, restore hidden parts of the layer (for example, for a girl on a green field, the system will automatically create layers "girl" and "field", and also speculates on what the grass and flowers, hidden behind the girl's silhouette, look like). With a few clicks, you can add grass and flowers to the background layer, change the girl's age, hair color, or replace her face with another girl's face.

7. The emergence of neural network models capable of removing various types of noise from a signal (images, sound, etc.) and improving its quality, learning only on noisy data.

Examples of using: restoration of old photos, videos, music records.

8. The emergence of neural network models capable of translating text by learning on monolingual corpora (that is, using a dictionary of one language).

Examples of using: improving the quality of machine translation without increasing the volume of bilingual corpora, decoding of undeciphered languages.

Internet and social trends

The collection of user data is more active

The fact that technology has become a necessary work tool and a way of leisure for many is no longer news. In 2018, the importance of the Internet has grown even more. Judging by the polls, few people are ready to part with it: 63% of Russian users want to stay online all the time. At the same time, in the past year, they began to talk about information security more often.

Many people voluntarily post information about themselves on the Internet, filling out profiles on social networks. Recently, the UK parliament has made public the documents of the company Facebook, which came to him during the lawsuit between the social network and the startup Six4Three. This information sheds light on how Facebook handles data. In particular, the company introduced the ability to receive information about calls and messages made by users from the Android platform. The update notice did not mention this. Subsequently, the text simply appeared in the People You May Know section.

Companies use user data to customize advertisements based on the demographics and geolocation of users. In 2019, these technologies will develop more strongly thanks to neural networks and other methods of working with large amounts of information.

Advertising messages will become even more personalized: they will try to penetrate more deeply into your personal life, reminding you which cafe you ate in yesterday, what products you bought in online stores and what you might be interested in in this regard.

If such attention seems excessive to you, it is worth taking action. At best, we receive intrusive advertising, at worst, we give potential attackers an opportunity to dispose of information in their own interests.

Online behavior is being monitored more closely

Today's social trends are closely related to technological ones, and the Internet is the main platform for discussing pressing social problems.

On job sites this year, offers have already appeared for specialists who search for and remove unwanted information on the Web. Until recently, this profession seemed fantastic. This is because each user leaves a digital footprint, which consists of the "cyberteness" of page views and active actions (posts, likes and comments).

At the same time, old actions on the Internet may turn out to be unacceptable from the point of view of the current legislation (in Russia, more than 85% of criminal cases about extremist statements fall on materials posted on the Internet) or the current agenda. For example, director James Gunn paid for immoral tweets eight years ago with a job at Disney. Regardless of how legitimate the punishments for reposts and obstruction due to old entries seem, it is obvious that today a page on a social network or microblog is not a personal, but a public space.

Social media tightening rules

Tolerance, diversity, intercultural communication and countering violence remain the main topics on the social agenda. Many brands of cosmetics and clothing rely on atypical models and abandon Photoshop in advertising photographs; in modernized versions of the series, the main characters become black (Buffy) and Latin American (Charmed), and the word of the year in 2018 according to the Oxford Dictionary became the adjective "toxic" - that is, poisonous and dangerous (for the environment, health or psychological comfort).

At the end of the year, global social media focused on becoming safe. Facebook has tightened the rules by banning the posting of content that "encourages sexual intercourse between adults."The ban includes invitations to shoot porn, massage or erotic dances, explicit sexual harassment, as well as "statements with erotic overtones", for example, "I want to have fun tonight." This is how the company fights human trafficking and harassment. According to representatives of the network, only the content that someone complains about will be removed.

In fact, the new Facebook rules prohibit not only harassment, but also any expression of sexuality.

However, the boundaries of the implicit erotic overtones are not clearly defined: it is not easy to come to a transparent agreement on the formalization of such things. By the way, the ban even applies to works of art, where people appear in sexual positions. This gives a lot of food for thought. How provocative, for example, are Michelangelo's David or photographer Terry Richardson's models?

Also, the publication of adult content was banned by the Tumblr service. Starting in December 2018, such images will be mercilessly blocked. However, the microblogging service focused specifically on pictures allows the placement of works of art and educational images. An exception is also made for images of breastfeeding, childbirth and images illustrating transgender transition.

Apparently, in 2019 we will have a lot of discussions about what is censorship on the Web and whether it should exist, and ethics commissions and specialized experts will have a lot of work to do.