The IDEAS NCBR’s working group that will deal with research in the field of computer vision will be headed by habilitated doctor engineer Tomasz Trzciński, professor at the Warsaw University of Technology and at the Jagiellonian University. The research agenda of the group will focus on issues related to the effectiveness of artificial intelligence models both in the context of the accuracy and pace of computations, as well as resources necessary for their operation.
Science and industry are increasingly relying on machine learning models, especially deep neural networks. The amount of data, and thus the parameters that models can take into account, increases. This translates into computing power needed for their analysis and, consequently, also energy consumption. It is estimated that the Internet alone consumes 70 billion kWh per year. Assuming that the IT sector will grow by 21% (including production and operations) until 2030, any possible energy savings will be worth their weight in gold. Due to its specificity, research in the field of sustainable development of AI is particularly important for computer vision because, for example, in the medical industry, where algorithms support medical personnel during surgeries using robots, the computational efficiency of the algorithms translates directly into the reaction time during the surgery and, thus, reduced potential for complications or shorter recovery time.
 Source: https://www.digitalinformationworld.com/2020/02/the-global-energy-consumption-of-information-technologies-infographic.html
“The research agenda of my working group concerns computational efficiency in a broad sense, in particular from the perspective of the reuse (recycling) of available resources. This is a direction that will allow algorithms not only to solve tasks but to do it effectively (quickly and accurately), using available data, infrastructure and previously completed calculations. Until now, to achieve this effect, scientists have focused on limiting parameters or computational memory taken into account by algorithms. “We decided to redefine the problem of the efficiency of machine learning algorithms and, as the first in the world, we will focus our research on the three pillars of efficient AI: the use of computation recycling, the use of available partial information and the accumulation of knowledge in scientists’ models in a continuous way,”
Research and development activities, as well as the education of a new generation of scientists are one of the main goals of the IDEAS NCBR. So far, groups have been launched under the leadership of habilitated doctor Piotr Sankowski, professor at the University of Warsaw (intelligent algorithms and data structures), and professor Stefan Dziembowski (digital currencies and smart contracts).
“I am glad that in such a short time we managed to establish as many as three research teams and attract such outstanding experts with international – not only scientific but also commercial – experience and achievements to the positions of leaders of the IDEAS NCBR’s research groups. Their recruitment is one of the biggest challenges we face, but it is also a key element of any research project. I believe that the team that we have managed to create will enable work on innovations that, in a few years, will find direct applications in the Polish economy” – commented habilitated doctor Piotr Sankowski, professor at the University of Warsaw, chair of the IDEAS NCBR.
“I am convinced that the research directions proposed by me will attract attention of talented young scientists from around the world, who will be able to combine work on the latest computer vision algorithms with activities for the sustainable development of technology, consistent with the zero waste concept.”. Currently, recruitment to my group for scientific and auxiliary positions, including the positions of team leaders, post-doctors or doctoral students, is underway. I also encourage all those who are just thinking about a doctorate or are graduates of computer science, mathematics and similar faculties to visit the IDEAS NCBR website and apply as part of the program prepared for doctoral students,” – added Tomasz Trzciński.
The full text of the press release is available here.