Self-learning neural networks are making great strides, delivering results that sometimes exceed human capabilities. This time, artificial intelligence has shown its superiority in the ability to notice the smallest details and draw conclusions on this basis.
This was proved in an experiment at the University of Minnesota, where neural networks gave the task to determine the age of a child by where his gaze is directed. As it turned out, a person's age betrays what his gaze is fixed on in the first place.
Before starting the experiment, data was collected to pre-train the AI. For this, the children participating in the study were divided into two groups. The first included participants at the age of one and a half years, the second - older children at the age of 2.5 years. With the help of a special mechanism that monitored the movements of their eyes, it was determined what children, one year apart, pay attention to in the first place. It turned out that the younger group first looks at the faces, while the older participants are more interested in objects, as a rule, those that can be reached.
All collected information became the basis for AI training. Later, the neural network was asked to indicate the age by the movement of the child's eyes. As a result, artificial intelligence technologies once again showed their abilities, in 80% of cases the algorithm gave the correct answer.
Such studies are not carried out as often as they involve small children. However, experiments of this type allow us to learn a lot of interesting things. The study helped to find out more details about human behavior. So, earlier it was assumed that the individual first of all fixes with his gaze those objects that are the most vivid and stand out more against the general background. It turned out that everything is not so simple. For a person, the meaning that this or that object carries is also important. Therefore, seeing something that excites him at the moment, a person may not notice brighter details.
In addition to age determination, AI technology is making significant progress in the music field. Thus, a project called Dadabots, run by two programmers, taught the neural network to compose music in the style of death metal. The developers maintain a Youtube channel where the algorithm they created showcases their creations. According to the creators of the project themselves, machine intelligence composes decent tracks for this musical direction without additional modifications and corrections.
To train their algorithm, the developers took as a basis the creativity of the Canadian collective Archspire, whose songs are distinguished by a high sounding rate. As a result, the artificial intelligence learned to create pieces in a "heavy" style, layering fast drums, guitar and aggressive vocals.
The creators of Dadabots talk about some progress of their "musician". Previously, when composing tracks in other genres, most of his works were rejected, and only 5% ended up in the final album of the fake group Dadabots. The current material does not require any modifications, so the developers gave the neural network maximum freedom, allowing it to compose music in stream mode.
According to the authors of Dadabots, improving the quality of music is related to the basis on which the AI was trained. So, the music of the Archspire group is fast, and the faster the drums sound, the more stable the neural network's music is. Previously, Dadabots have already released many compilations of various genres, including an album in the style of the Beatles.
The Topic of Article: Artificial intelligence learned to determine age by eyes and compose music in a ”heavy” style.