Neonka from Sberbank – Kommersant

German Gref’s contract as head of Sberbank can be extended for another four years, that is, until November 2027, Frank Media reported on Wednesday, May 10, citing its sources. And this is despite the active rumors at the beginning of the year that the head of Sberbank may lose his post. Finance Minister Anton Siluanov and Yandex corporate development adviser Alexei Kudrin were named as likely successors.

But against the backdrop of the results of Sberbank, which, having fallen under sanctions, was able to show a profit of 300 billion rubles at the end of 2022. (most of all the banking sector), and this year he planned to increase it four times, retaining the current leader looks absolutely logical.

Much more interesting is the remark of one of the agency’s sources that the bank’s new strategy will focus on artificial intelligence (AI). It’s no secret that this is one of the bank’s favorite topics. By December 2022, 99% of all decisions on loans to individuals there are made by AI. In addition, AI-based solutions are being developed for industry, medicine, education, and legal services. At the end of April, Sberbank presented its own version of the GigaChat multimodal neural network.

Meanwhile, abroad, where the race for AI began a little earlier than in the Russian Federation and moved faster, recently there have been more and more voices for limiting this direction. At the end of March, an open letter even appeared, signed by Elon Musk, Steve Wozniak (one of the founders of Apple) and more than a thousand respected people, demanding to stop developments in the field of artificial intelligence. Its authors seriously fear that AI can take over the world.

Whether the neural network of Sberbank will pose a threat to humanity in the next four years of its leader’s work on the development of AI is difficult to say. But for now, there doesn’t seem to be anything to worry about. As follows from the description, “the architecture of the GigaChat service is based on the neural network ensemble of the NeONKA (NEural Omnimodal Network with Knowledge-Awareness) model… Thanks to this, the Sber neural network can… support a conversation, write texts, and answer factual questions.”

By their name, the authors of the model refer to the heuristic machine described by the Strugatsky brothers in the 1968 story “The Tale of the Troika”. Inventor Edelweiss Mashkin presented it as “an accurate electronic-mechanical device for answering any questions.” The car was tested on the question of what kind of light bulb it has inside. “I have a neon inside me,” the device answered confidently. There was only one nuance in this dialogue – the machine was “not fully automated.” The inventor himself printed the questions for her and her answers. If Sberbank’s NeONKA is built on the principles of continuity with the Strugatskys’ “neonka”, humanity, apparently, can not be afraid of this AI.