< br />
MOSCOW, August 15 AI assistants in smartphones in the next five years will begin to understand users at a glance and will give advice based on the memory of their actions, said vice — MTS President for Strategy and Innovation Evgeny Chereshnev.
“In the next five years, we will see generative AI begin to first take into account a broader context when generating answers: simply put, if you ask it to recommend where to go on vacation, the AI agent will take into account your financial capabilities, past vacation experience, your preferences in terms of the type of recreation and climate and will give the most relevant answer. The main direction of technology development is to reduce the user’s efforts when creating a request; artificial intelligence will have to understand it perfectly,” said Chereshnev. 
Smartphones, in turn, will transform from a «device-brick with a set of icons that you can poke at» into a digital assistant with a memory of the user's actions. This development scenario will allow each user to have their own personal AI agent, which will live not only in the smartphone, but also in the TV, smart speaker and a dozen other devices at home.
«And in the end, these AI assistants will be of two types. In the future, the memory of some will be entirely owned by corporations, and the memory of others — by the user. The closest analogy here is the cloud business. One infrastructure is deployed on the provider's side, and the other — on the premises of the data owner,» the expert added.
Chereshnev also spoke about the risks that the development of such technologies poses for users. One of them is that generative AI, which will operate on board millions of smartphones around the world, will be able to receive and analyze large amounts of data about each user in real time — for example, how much they earn and where they spend money, who they meet with and when. And the owner of the technology will be able to manage the information at their own discretion — more precisely, target advertising or push a person to certain actions.
Corporations will also receive data from the devices of people united by a certain characteristic — this could be a place of work, service or residence. There is a risk that such AI, having analyzed and compared data from thousands of smartphones of employees of a closed enterprise, will be able to find out closed data about it. «This opens up broad opportunities for both manipulating the opinions of social groups and individual decision-makers,» the expert explained.
«Such risks can only be avoided in a scenario where AI is deployed not on the infrastructure of an international corporation, but in a secure cloud on the user's side,» Chereshnev added.