User Interface Professors Discuss Artificial Intelligence and Computer Vision

UI professors discuss artificial intelligence, computer vision

Machines learn.

As more and more of today’s world is automated for maximum efficiency, computer software has the potential to perform tasks usually reserved for humans.

Artificial intelligence is the ability of a computer or computer-controlled robot to perform tasks that are typically performed by humans, as these tasks require human intelligence and judgment. Computer vision is a subset of AI, in which computer systems use image capture software to understand and interpret the visual world. Computer science professors at the University of Illinois at Urbana-Champaign are using AI and computer vision both in the classroom and outside.

AI is becoming more and more common in the contemporary technological landscape. However, misconceptions on the ground remain. Dr. David Forsyth, professor of engineering at the University of Illinois at Urbana-Champaign, noted that despite recent startup fervor around AI and computer vision, most systems applications are less flashy and practical.

“Everyone and their cat is here to start AI businesses and it mostly comes down to a classifier,” Forsyth said. “Computer vision involves doing useful things with images…for example, surveillance of airports, you take footage of people picking up bags where air traffic is coming in, and then you use automated methods to track the bags and make sure the bag leaves customs with the same person who picked it up.

Building an AI system can use machine learning, a data analysis method that automates analytical models. However, to start a system, data must be captured from human input. Business professor Dr. Justin Leiby noted that some machine learning algorithms use groups of online workers to identify images, thereby training certain algorithms. When an individual identifies a certain image, the algorithm takes note of it and begins to make connections between the visual input and the individual’s classification.

“It’s a tool for learning how to teach algorithms,” Leiby said. “It’s sort of their training device. Human beings are able to (classify pictures), and you can start mimicking human classification if you have enough people to do it. So ask a few thousand people to answer a question, show them a picture of a chair, and say, “Which room in the house does this chair belong to?” And suddenly, because thousands of people are telling an algorithm how to recognize a chair that fits in the living room versus a chair that fits in an office from its photo, the algorithm begins to learn how to make its best guess. .

Engineering professor Dr. Derek Hoiem has spent most of his career studying how AI and computer vision can make the world a better place. Recently, Hoiem has taught courses such as CS 445: Computational Photography, CS 598: 3D Vision, and CS: 543 Computer Vision to undergraduate and graduate students.

As AI research develops, systems have become more sophisticated and adept at analyzing a wide range of data. Yet, there is still room for improvement. One of Hoiem’s ​​current research directions focuses on general-purpose vision — a way to make an AI system more capable of solving a wide range of tasks instead of focusing intensely on a singular specific task.

“So, for example, like with people, we can see and we have hands and we move our bodies to gather information,” Hoiem said. “So we can do all kinds of tests in line with our senses and the actions we can take, and we can potentially learn to do the task and do it well. The same idea is applied to general purpose vision, the idea is that within the scope of what the AI ​​system can see and what it can do, it can perform any task that the scope allow. So it can learn to detect objects, or it can learn to caption images or classify images.

Hoiem doesn’t just focus on AI and computer vision from an academic perspective. Hoiem also works as director of strategy for Reconstruct – a company using computer vision to map construction sites and provide real-time updates to on-site contractors. Hoiem, along with colleagues Mani Golparvar and Tim Brettell, founded Reconstruct in 2015 to apply computer vision to solve a long-standing construction problem: delays.

“It’s a very important way for computer vision to benefit society,” Hoiem said. “Building and construction infrastructure underpins our economy, and there is potential to meet this critical need using the ability to create 3D models from photographs and do reconnaissance. It is well known that “Construction often gets way behind schedule and over budget. Our idea was that we could use computer vision to provide that situational awareness on construction sites.”

Ethics is at the forefront of the contemporary conversation about AI and computer vision. With the technology already established, researchers and activists are beginning to discuss the implications of creating and using AI.

In 2019, a recent AI project called Speech2Face – which used machine learning to create an algorithm designed to develop facial images based on recording speech – came under fire after sociologists and computer scientists raised concerns ethical implications of the project.

When using machine learning, systems are only as good as the data received as input. Although the computer system itself is not biased, the information it receives is still determined by a human sensitive to its internal biases. Examples include facial recognition software trained only to recognize mostly light-skinned men, failing to recognize those with darker skin.

Hoiem noted that one of the main ethical issues stems from how developers’ input of information into systems can be biased, whether that bias is explicit or implicit.

“Ethically, where it really creates problems is that sometimes AI researchers and companies will create a product that will be developed based on a certain distribution of images or data,” Hoiem said. “For example, if it’s trying to identify faces, that may be based on company employees and other subjects they may have been able to obtain data for, but that may not represent the broader demographics of the user base or environment it will be applied to.

[email protected]

[email protected]