Innovation

Currently, one of the most important issues facing humanity is emissions of carbon dioxide because, it makes an active contribution to climate change and environmental pollution. At the moment, many countries are trying to minimize the negative effects of carbon on the environment.

According to recent estimates, by 2030, increased use of Artificial Intelligence (AI) will increase electricity consumption in Europe by three times. Currently, the entire European energy supply system is facing serious problems in meeting this growing demand.

These forecasts are related to the continuously growing trend towards the use of AI in specialized tasks and everyday life. Each AI that is available to ammass user pool, uses data center resources that consume energy. According to scientists, by the end of the decade, data centers will consume about 150 TWh of energy, which is more than 50% higher than the current consumption of data centers.

The solution to this problem lies precisely in reducing the energy that the data center emits when the AI is running. This is the problem that Smart Engines tackles, when developing the Green Optical Character Recognition (OCR) product.

Green OCR is a data recognition technology based on environmentally friendly and responsible AI. Green OCR technology has been developed in accordance with the Doctrine of Responsible Artificial Intelligence. It is based on three basic principles: environmental friendliness, security, and privacy. Green OCR is used to recognize ID documents (ID cards, passport, driving licence, etc.), MRZ (Machine Readable Zone), bank cards, barcodes, as well as documents with and without fixed forms in the video stream, photos and scans.

As our interviewee Ekaterina, PR Director of Smart Engines, said about Green OCR:

“We were able to make a technology that, while maintaining the quality and speed of recognition, consumes much less energy and, accordingly, the environmental impact is also less, by about 3-4 orders of magnitude, i.e. by 30-40 times.”

These results are achieved, since calculations do not require significant computing resources and the use of additional accelerators in the form of energy-intensive graphics processing units (GPU).

Inspiration

The company was initially focused on intensive rather than extensive development growth. Therefore, they have been interested in issues of responsible production since the creation of Smart Engines.

“Our scientists were thinking about how to work without additional computing power, how to optimize calculations. And they just thought, how can we solve the problem from this point of view, so that we can avoid resorting to GPU and maintain computational efficiency? So that's what we did”, said Ekaterina.

It was a gradual path of development. Step by step, the company’s scientists have optimized calculations to provide better and more affordable recognition services. When the highest computational efficiency was achieved, the company measured the carbon footprint before and after using the new technology. Having received such significant results, Smart Engines realized that it is able to contribute to the preservation of the well-being of the planet by following its own methodology by creating responsible and environmentally friendly AI.

Overall impact

Smart Engines became one of the first companies from the industry to join the United Nations Global Compact in 2020. The introduction of Green OCR AI technology into all the company's products makes it possible to process documents faster and more accurately and at the same time significantly reduce energy consumption during recognition, which on a planetary scale reduces the negative impact on the environment.

The company is one of the trendsetters of energy-efficient AI, regularly publishing scientific articles on this topic and delivering talks at the international conferences. Thus, throughout the year of 2023, the company’s scientists participated in three international conferences: the 17th International Conference on Document Analysis and Recognition (ICDAR 2023), the London International Meeting (LIM 2023), and the 16th International Conference on Machine Vision (ICMV 2023). Smart Engines scientists presented more than 20 talks at these conferences in total.

“When we calculated the difference in the carbon footprint, which was at the beginning and is now, we already started talking about it. Other companies are guided by us in this sense”, said Ekaterina.

In addition to minimizing the negative impact on the environment, the company pays special attention to its employees.

“Our company is human-centered: we take care of our employees so that everyone feels good”, said Ekaterina

Business benefit

The use of energy-efficient algorithms makes it possible to use high-tech solutions even on devices with limited computing power. Due to this, the company has new customers. Over the past year alone, the company's solutions have been used by more than 275 million people worldwide, and this figure is steadily growing.

The advantage of using less computing power is complemented by a positive impact on the environment. Thanks to the introduction of Green OCR technologies and the company's active work to popularize energy-efficient artificial intelligence, the business has not only reduced the carbon footprint of its products, but also strengthened its reputation in the market as a leader in environmentally friendly innovations. These steps have led to increased awareness among stakeholders, increased customer base and attracted new partners interested in sustainable development. This contributes to the formation of an expert community focused on sustainable development, strengthens the reputation of Smart Engines as a leader in the field of environmentally friendly AI and attracts new partners interested in joint innovative projects.

Watch video on YouTube

Social and environmental benefit

The use of artificial intelligence, namely computer vision, makes it possible not only to simplify the process of filling out documents for the client, but also to completely convert the entire document flow into an electronic format. The importance of this optimization for society is quite obvious. Through the use of artificial intelligence, we get a new level of service, increase labor productivity, as we negate routine operations.

The transition to electronic document management also allows us to preserve nature, since we no longer need to cut down forests for paper production. In addition, optimization of neural network learning algorithms has allowed the company to reduce carbon dioxide emissions. Compared with similar learning models, the considered solution spent 3-4 orders of magnitude less energy. This solution is the best in both the short and long term. Currently, in companies offering AI services, as the complexity of the model increases and the number of parameters increases, the cost of ensuring the smooth operation of the datacenter and the amount of hardware itself increases exponentially. And the leaders of the AI solutions market, instead of optimizing the existing approach, resort to vertical/or horizontal scaling, as evidenced by OpenAI's plans to build a new data center, which requires a small nuclear reactor to provide the necessary amount of electricity. That is why this approach is unique in the industry. It allows you to reduce costs in the present and reduce the negative impact on the environment in the future.

Interview

Ekaterina Remizova, PR Director