Google Research VP Jeff Dean identified five top Machine Learning trends that will have a big impact on society.

1. More powerful general-purpose ML models

Jeff Dean, Senior Fellow, and VP of Google Research published a Google study on trends in machine learning in January 2022. It brings together the main areas of Machine Learning (ML) that will have the greatest impact on people’s lives.

Researchers are currently working with larger and more powerful machine learning models. In addition, the growth in the amount of data and the size of the models themselves (for example, it is now possible to generate models with hundreds of billions of parameters) has led to an increase in the accuracy of performing various tasks. Along with single-modal models, for example, which recognizes an image, speech, or sound, multimodal models begin to appear – they are able not only to understand different modalities but also to give results in them.

Jeff Dean sees this as an opportunity to create high-performance, general-purpose models that can process different kinds of data and solve thousands or millions of problems. In the near future, Google plans to implement this vision in a new artificial intelligence architecture – Pathways, which will combine many of the ideas that have been developed independently until now.

2. Improving the efficiency of machine learning

Against the backdrop of the development of computer hardware, the improvement of machine learning algorithms, as well as research in the field of meta-learning, the capabilities of ML models are rapidly expanding. In addition, compared to previous years, it turns out to spend much fewer resources on every aspect of machine learning. Increasing the efficiency of machine learning also allows you to develop even larger, better, and more affordable models at a lower cost.

According to the VP of Google Research, to improve the efficiency of models, it is necessary to work on improving the performance of ML accelerators, improving the compilation and optimization of work tasks, as well as introducing sparseness, which means that when training or performing a certain task, only the most appropriate part of the model is activated.

3. Machine learning is becoming more and more useful for both the individual and society

Machine learning innovations have led to new developments that are beneficial to both the individual and society as a whole. Jeff Dean notes that new developments have improved features such as mobile photography, live translation, and so on. It is worth noting that two trends are observed in this area: the personalization of technological advances and the strengthening of privacy measures.

The last point is especially important – at present, the issue of cybersecurity has become the most relevant. ML technologies have brought features such as suspicious message alerts and safe navigation, which offer alternative routes in the virtual space.

It is impossible to ignore the various areas of business. Today, ML helps to receive and analyze information much faster, which optimizes internal processes, and reduces the cost and risks of initial investments, increasing the bottom line for companies. In today’s marketplace, you can already choose the right enterprise analytics platform to implement data pipelines and consistently get insights at scale. In this way, business companies can turn a profit 10 times faster and focus on what matters to the business: business intelligence, data science, and data-driven decision-making through machine learning.

4. The growing impact of machine learning on science, health, and the environment

The influence of machine learning on the development of a number of fundamental sciences is growing rapidly. With the help of ML, in the near future, it will be possible to solve a wide range of problems in various fields – from renewable energy to medicine.

Among the cutting-edge developments based on machine learning is computer vision, which is used both to reconstruct the neural structure of brain tissue and to predict the weather. In addition, the method of automatic simulation shows great promise. Jeff Dean also praises the potential for machine learning to be applied to healthcare and the global climate crisis.

5. Deeper and broader understanding of AI

Google Research considers it necessary to develop new methods to ensure the fair and equitable use of artificial intelligence in society. To do this, researchers are studying how the various components of ML work together and separately. According to the head of this direction, the creation of publicly available comprehensive and less biased databases is an opportunity to improve machine learning for everyone.