AI Worse at Recognizing Images than Humans
Researchers from HSE University and Moscow Polytechnic University have discovered that AI models are unable to represent features of human vision due to a lack of tight coupling with the respective physiology, so they are worse at recognizing images. The results of the study were published in the Proceedings of the Seventh International Congress on Information and Communication Technology.
To understand how machine perception of images differs from human perception, Russian scientists uploaded images of classical visual illusions to the IBM Watson Visual Recognition online service. Most of them were geometric silhouettes, partially hidden by geometric shapes of the background colour. The system tried to determine the nature of the image and indicated the degree of certainty in its response.
It turned out that artificial intelligence is not able to recognize any imaginary figure, with the exception of a coloured imaginary triangle. Due to the high contrast with the background, it was recognized correctly.
‘Objects similar to those that we used during the experiment can be found in real life,’ says Vladimir Vinnikov, an analyst at the Laboratory of Methods for Big Data Analysis of HSE Faculty of Computer Science and author of the study. ‘For example, autopilot of a car or airplane perceives a trailer or a radio tower, which at night are indicated only by marker lights, the same way as we perceive imaginary geometric shapes.’
The human eye is constantly moving involuntarily, and the photosensitive surface of its retina has the shape of a hemisphere. A person can see an illusion if the image is a vector, i.e., if it includes reference points and curves connecting them. The human imagination will complete the picture due to constant eye movement, a physiological feature of our vision.
In optoelectronic systems everything is arranged differently. Their light-sensitive matrix has a flat, usually rectangular shape, and the lens system itself is not nearly as free in movement as the human eye. Therefore, artificial intelligence cannot complete imaginary lines that connect fragments of a geometric illusion. Machine vision sees only what is actually depicted, whereas people complete the image in their imagination based on its outlines.
Today, neural network image recognition systems are actively spreading in the commercial sector. However, the question of how accurately machines recognize images is still open. Human lives may depend on the accuracy of recognition. For example, an accident may occur if the autopilot of a car or airplane does not recognize an object with low contrast relative to the background and is not able to dodge an obstacle in time.
Scientists believe that inaccuracy of machine image recognition can be corrected. For example, they can complement the recognition of raster images, which represent a grid of pixels, by simulating physiological features of eye movement that allow the eye to see two-dimensional and three-dimensional scenes. An alternative way is to add vector description of the images, which will help to programme the machine to bypass the image along the trajectories specified by the vectors.
‘Imaginary objects should definitely be used as tests in systems that depend on the recognition of photo and video streams, for example, in autopilots of cars or drones. This will help to avoid the risks associated with the use of machine intelligence systems in industry and transport systems,’ believes Vladimir Vinnikov.
The SNAD team, an international network of researchers including Matvey Kornilov, Associate Professor of the HSE University Faculty of Physics, has discovered 11 previously undetected space anomalies, seven of which are supernova candidates. The researchers analysed digital images of the Northern sky taken in 2018 using a k-D tree to detect anomalies through the ‘nearest neighbour’ method. Machine learning algorithms helped automate the search. The paper is published in New Astronomy.
Tharaa Ali, from Syria, got her Master’s in Moscow and is now a first-year PhD student in informatics and computer engineering at HSE University. Below, she talks about the use of AI in learning, winning scholarships, and doing research at HSE University.
Whether researching how the human brain works, identifying the source of COVID-19, running complex calculations or testing scientific hypotheses, supercomputers can help us solve the most complex tasks. One of the most powerful supercomputers in the CIS is cHARISMa, which is now in its third year of operation at HSE University.Pavel Kostenetskiy, Head of the HSE UniversitySupercomputer Modeling Unit, talks about how the supercomputer works and what kind of projects it works on.
The HSE Centre for Artificial Intelligence, together with its partners in industry, is working on 25 applied projects in the fields of telecommunications, finance, education, medicine, etc. The results of the work by researchers and developers were recently presented at a meeting of a Russian government working group. That meeting summed up the initial results of the federal Artificial Intelligence project, part of the national Digital Economy programme.
A research team from the HSE University Artificial Intelligence Centre led by Ivan Yamshchikov has developed a model to predict the success of efforts to rehabilitate homeless people. The model can predict the effectiveness of the work of organisations for the homeless with about 80% accuracy. The project was presented at a conference dedicated to the activities of social centres.
The Federal Brain and Neural Technology Centre at the Federal Medical and Biological Agency is launching the Laboratory of Medical Neural Interfaces and Artificial Intelligence for Clinical Applications, which has been created by employees of HSE University. Read below to find out about the Laboratory and its objectives.
In Mexico, a pilot project applying artificial intelligence (AI) algorithms enabled the Tax Administration Service to detect 1200 tax-evading companies and 3500 fraudulent transactions within three months – a task that would have taken 18 months using conventional methods. Despite some obvious benefits, the use of AI-based solutions to counter corruption also entails several challenges, according to experts of the HSE Laboratory for Anti-Corruption Policy (LAP) and the HSE Faculty of Law who have examined the relevant experience of several countries. A report based on the study’s* findings was presented at the XXIII Yasin (April) International Academic Conference hosted by the Higher School of Economics.
Innopolis University has announced the results of Global Al Challenge, an international AI industry online hackathon in which teams of developers compete to create new materials using artificial intelligence. The DrugANNs team, which included students from the HSE University Faculty of Computer Science, took third place.
In recent years, advanced technologies for creating deepfake images have made it almost impossible to distinguish them from real photos and videos. Researchers discussed the future development of deepfakes and how to protect yourself from new types of fraud during the round-table discussion ‘Fake News: An Interdisciplinary Approach’ as part of the XXIII Yasin (April) International Academic Conference on Economic and Social Development.
The transformation of all aspects of economic and public life due to new technologies necessitates the development of a legal framework for the application of innovation. The HSE Faculty of Law has hosted the 10th international scientific-practical conference ‘Law in the Digital Era’ online. The participants included a wide range of experts from Russian and international universities, public bodies and other organisations.