7 advances in artificial intelligence in 2018

1. Painted an expensive picture

The golden baguette, print on canvas and the formula instead of the artist's signature in the corner is "Portrait of Edmond Belamy" painted by artificial intelligence. Christie’s became the first auction house to sell Is artificial intelligence set to become art’s next medium? an AI painting: the canvas sold for $ 432,500 with a starting price of $ 7,000.

A team of French specialists from the Obvious company worked on the project. They used a generative adversarial network algorithm that consists of two neural networks. The first, the generator, looked through 15 thousand portraits from the 14th to the 20th centuries and created her own image based on them. The second network, the discriminator, compared the work of the generator to pictures that people painted. The result was considered successful when the second network did not find the difference between the real web and the operation of the generator.

2. Created a horror movie

It turned out to be a strange and frightening black and white short film with incomprehensible dialogues. An artificial intelligence named Benjamin became the screenwriter, director and editor of the tape. First, he came up with a plot, then he found the necessary frames from old films, edited them in the right order and superimposed the faces of modern actors on the heroes of black and white cinema.

The creators of Benjamin say that not everything went smoothly: mustaches sometimes appear on the faces of women, and some remarks do not always fit the meaning. They attribute this to the fact that Benjamin had only 48 hours to work.

3. Learned to determine age by eyes

You can hide your age from anyone, but not from the AI. A group of scientists from Moscow State University, together with the technology startup HautAI OU, has created an artificial intelligence PhotoAgeClock that can determine PhotoAgeClock: deep learning algorithms for development of non-invasive visual biomarkers of aging the chronological age of a person in the eyes.

The area around the eyes is not very susceptible to the influence of lifestyle or other factors and changes naturally with age, so the team of scientists chose this area for analysis. The neural network studied 8,500 photographs of the area around the eyes and learned to determine age with an accuracy of two years.

4. Started developing drugs

It takes decades to develop drugs, and then a few more years to laboratory and clinical trials. This is a painstaking and long process.

At the University of North Carolina School of Pharmacy, scientists have created an Artificial intelligence system created at UNC-Chapel Hill designs drugs from scratch AI from two neural networks. One is loaded with data on the structure and properties of molecules, as well as the desired effect. The second neural network learns from the first: it assimilates this data and selects possible solutions. AI is now working with over 1.7 million molecules. This will help to significantly speed up the process of developing new drugs, and successful results can become the basis for the creation of, for example, new antibiotics.

5. Helped the Vatican

There are many legends about the Vatican archives. But what is known for sure, in 2017, the In Codice Ratio team received a request from the state to study the texts of the archives and digitize them using AI. The first results began to appear only in 2018. The total length of the shelves in the archive is 85.2 km, and now only a few millimeters of the collection have been decoded and posted on the Web.

The process turned out to be more complicated than expected. Optical Character Recognition (OCR) technology has long been used for text in books and other printed documents. But the materials from the Vatican archives are ancient manuscripts written in different fonts and in different languages. OCR does not always understand where one letter ends and another begins. As a result, the researchers tuned OCR not to recognize letters or words, but to the details of the face. The algorithm learns lines and lines and assembles them into possible variants, like a puzzle.

6. Released a fashionable collection

Last week of fashion in Moscow, Ksenia Bezuglova, activist and Miss World 2013 among girls in wheelchairs, presented the Fashion of the Future: Cyber Vision, Robots and Artificial Intelligence at the Mercedes-Benz I-INCLUSIVE I-INCLUSIVE collection. She worked together with leading technology centers in Russia. Together they showed clothes and accessories of the future for people with disabilities.

For example, the work of Ksenia and the Sensor-Tech laboratory is a smart cane with AI for the blind and deaf-blind. The cane can recognize faces, objects and obstacles, measure the distance to them. The aim of the project was to help people with disabilities lead a more independent lifestyle and get to know the world of fashion.

7. Became a news anchor

In China, artificial intelligence has become China's state-run Xinhua News Agency debuts ‘AI anchor’ to read the news a popular presenter this year. He reads the news in English and looks and sounds like a real person - Zhang Zhao, an employee of the Xinhua News Agency.

The simulation was entirely generated on a computer: they combined the texts of the speakers, facial expressions and lip movements of real people. A video with a real person is loaded into the program, and AI, using machine learning, independently analyzes gestures, conversation patterns and other details in order to further reproduce it on the air.

The field of AI is gaining momentum more and more. Probably, next year they will launch projects that are now in test mode: unmanned taxis, AI-controlled submarines, or virtual border guards who conduct inspections at airports. We will follow the development of technology with interest, will you?