2 Mar '17
An international research team has come up with a self-learning artificial intelligence, a breakthrough that may have wide applications, including therapies to treat human memory dysfunctions. The key participant in the project is the Tomsk State University (TSU). Other developers also include researchers from Germany, Bulgaria, Ukraine, Belarus, and Kazakhstan.
As Prof. Vladimir Syryamkin of a TSU laboratory explained, “a mathematical and computer model of the human brain was built first, followed by the development of a radio electronic device that contains perceptrons [in machine learning, an algorithm for supervised learning of binary classifiers—Editor’s note]. The device is capable of processing an array of information, including video, sounds, etc. Work is currently under way to develop the core of our robotic complex, which is a smart control center.”
“At the end of the day, the artificial brain must become an analog of the biological model. We have a lot of work ahead, but a very important step has been made, which is the unraveling of the mystery of the cerebral neural network. In our physical model, as is in the actual human brain, new neural links keep forming and old ones keep dying. In the human body, this is the process of forgetting things,” commented Vladimir Shumilov, a Ukrainian researcher from Kiev who was the chief developer of the project and also a research fellow at TSU’s Computer Vision Systems lab.
According to the scientists, their physical model has the ability to train itself and accumulate ‘life experience.’ The artificial carrier of intelligence can react to external signals such as light or sound and is reportedly capable of looking, by trial and error, for solutions that would help it avoid being impacted by these ‘irritants.’ Until the right decision is found, neurons (perceptrons) are in excitation; with the right decision found, the artificial intelligence is expected to memorize it and then use it in similar situations.
Prof. Syryamkin said that “very soon specialists from other fields, including biology and psychology, will join the effort. It’s not only thinking and analyzing all on its own that the artificial intelligence is expected to be capable of; it is also expected to tell good from evil from the ethical standpoint.”
Applications for the new type of artificial intelligence appear to be “virtually unlimited,” the researchers think.
“Our solution might help… in research and treatment of a variety of amnesias, and also of the Alzheimer’s and Parkinson’s diseases, as well as other dementias. All these disorders typically have the same root cause, which is either damaged neural links or neural activity decrement. Physicians could use an artificial brain to model pathologies, and then draw on the experience to choose drug-based therapies to address the problems,” Prof, Syryamkin said.
In addition, the artificial carrier of intelligence developed in the Tomsk-based international effort could become part of smart robotic systems and neurocomputers, the project owners feel.