From Invisible to Indispensable: The Integration of Machine Learning in Daily Life and Education

In the near future, machine learning will be ubiquitous in technology, and everyone will be able to use it, not just experts. It's already challenging to find someone who hasn't used a product or service powered by this technology. I believe that soon people will use machine learning intentionally, and the basics of machine learning will be taught and studied in schools and universities in the next 10-15 years, with its benefits significantly contributing to further education.

Reading Time: 4 minutes

petr ermakov

Illustration: Lenka T.

Artificial Intelligence (AI) and machine learning (ML) have been in the spotlight since OpenAI launched ChatGPT. However, AI and ML are much older than ChatGPT; the first research on artificial intelligence dates back to the 1950s. Also, AI and ML are much more than the increasingly prevalent chatbots. 

Petr Ermarkov, ML Brand Director at Yandex, discusses what machine learning is, its presence, and how it will impact our future. 

WebMind: How would you define machine learning in today’s technological environment and its influence on various industries? 

Petr Ermarkov: Machine learning is considered one of the primary drivers of technological progress. Its impact extends through numerous technical products and services, shaping their functionalities. From smart virtual assistants that understand and respond to personalized recommendations in e-commerce to the technological wonders, ML provides the chance to learn, adapt to humans, and make daily life easier. 

WebMind: How do you approach the process of integrating machine learning models into applications used in our daily lives? 

Petr Ermarkov: Yandex has a long tradition of using machine learning. For instance, since 2009, when the world began talking about this phenomenon, we have developed and implemented our own method of machine learning called MatrixNet. One important characteristic of this method is its resistance to overfitting, which is crucial given the scale at which we operate. Building on that, we created a more advanced technology – CatBoost. With it, we build our own ranking formula that powers our internet search engine. 

In general, we have numerous programs based on ML technology that we use through mobile applications. As another example, I would mention SpeechKit – a speech recognition technology that lays the foundation for live video broadcasting. Our recommendation algorithm helps predict the music users would like to hear and also represents a certain product in our market that users might want to buy. 

Integrating machine learning into applications not only helps companies be more efficient but also provides a significant competitive advantage as it relies on the latest neural networks. 

To use neural networks, you need highly skilled specialists working with data. As early as 2007, we noticed a shortage of qualified personnel in this field, so we started a data analysis school within Yandex. Thanks to the two-year program of this school, we managed to educate over 1000 top-notch specialists. 

WebMind: In which direction will machine learning continue to develop, and how will it impact social life? 

Petr Ermarkov: In the near future, machine learning will be omnipresent in technology, and everyone will be able to use it, not just experts. It is already challenging to find someone who hasn’t used a product or service powered by this technology. I believe that people will soon intentionally use machine learning, and in the next 10-15 years, the fundamentals of machine learning will be taught and studied in schools and universities. I consider that its usefulness and significance will significantly contribute to further education. 

WebMind: What are some of the most significant advances and discoveries in machine learning that you have recently noticed? 

Petr Ermarkov: One area that has particularly caught our attention is generative models. They have truly revolutionized the way we use technology. For instance, generative language models have an incredible ability to create content that appears natural and contextually relevant. This kind of technology drives chatbots where you feel like you’re talking to a real person. 

When we started working with ChatGPT last year, we realized the impact of such language models. On that occasion, we decided to take it a step further within the company and created our own language model, YandexGPT, which became our voice assistant, Alice. 

In addition to language models, we are also working on our neural network for generating images called Shedevrum. With simple input of descriptions, images can be generated based on your request. 

WebMind: How does Yandex implement machine learning into its own products and services? 

Petr Ermarkov: From the very beginning, we have integrated AI and ML technologies into our products, and now they are an integral part of our services and products. It’s hard to find an example of a product or service that is not based on machine learning. Whether it’s ride-hailing services, delivery, maps, search, e-commerce, or ads, ML technology is a standard component of these and many other services. 

Let’s take Yandex Go app as an example. The algorithms used in this app help efficiently distribute orders by predicting the most efficient tasks. If I were to describe each of the over 90 services we have, this would be a very long interview. However, ML algorithms are an essential part of our products for more efficient functioning of the application. 

WebMind: How do you ensure that the machine learning solutions developed by Yandex are fair and objective? 

Petr Ermarkov: Machine learning is simply a tool used within programming. Neural networks analyze vast amounts of data, and that data is created by people with different opinions, skills, or knowledge. On the other hand, neural networks do not have their own opinions; they merely repeat what has been previously learned from other developers. In Yandex, we take all necessary measures to ensure the integrity of data and, at the same time, monitor the predictions of the network to detect potential “harm” on time. 

In support of this, we have created a new profession – AI trainers – who analyze texts used to train neural networks. AI trainers are experts in finding information and fact-checking. This ensures that our neural networks provide honest and inherently objective responses. 

"Ever tried. Ever failed. Never mind. Try again. Fail better."

Subscribe to our newsletter and stay updated !