Soon, we will no longer have to use the Help function. The computer will detect that we have a problem and come to the rescue itself. This is one of the possible implications of new research at the University of Copenhagen and the University of Helsinki.
“We can make a computer that will edit images entirely based on human-generated thoughts. The computer has absolutely no prior information about what parts need to be edited or how. It has never done this before, ”said Associate Professor Tuukka Ruotsalo, Department of Computer Science, University of Copenhagen.
The results are presented in an article accepted for publication at CVPR 2022 (Computer Vision and Pattern Recognition), the most renowned conference in the field.
Brain activity as the sole input
In the accompanying study, 30 participants were equipped with hoods with electrodes that mapped electrical brain signals (electro-encephalo-graphy; EEG). All participants were given the same 200 different face images to view. Also, they are given a series of tasks like finding woman’s faces, finding the elderly, finding blond hair and so on.
Participants do not take any action, just look at the pictures briefly – 0.5 seconds for each picture. Based on their brain activity, the machine first maps to the given desired and then edits the images accordingly. So, if the task is to find the elderly, the computer will change the images of the young, making them older. And if the task is to search for a given hair color, everyone will get that color.
“In particular, the computer has no knowledge of face recognition and has no idea about gender, hair color, or any related features. However, it only edits the part in question, left other parts of the face unchanged, ”commented PhD Student Keith Davis, University of Helsinki.
Some might argue that a lot of software capable of manipulating surface parts is already available. That’s not the point, Keith Davis explains:
“All existing software was previously trained to have an input label. So, if you want an app that makes people look older, you feed it thousands of photos and tell computer where the children are, and how old they are.Here, the brain activity of the subjects is the only input.This is a new paradigm of artificial intelligence – which uses the human brain directly as the source of input .
Possible medical applications
A possible application may be in medicine:
“Doctors are already using artificial intelligence to interpret the scan images. However, mistakes do happen. After all, doctors are only helped by the images but they themselves make the decisions. Maybe some parts of the images are more often misunderstood than others. Such patterns can be discovered through the application of our research, ”said Tuukka Ruotsalo.
Another application could help some groups of the disabled, for example allowing a paralyzed person to operate his or her computer.
“That’s good,” commented Tuukka Ruotsalo, adding:
“However, that is not the focus of our research. We have a wide scope, looking to improve machine learning in general. The range of possible applications could be wide. For example, 10 or 20 years from now we may no longer have to use a mouse or type commands to operate our computer. Maybe we should just use mind control! ”
Call for policy regulation
However, the coin has a broken side according to Tuukka Ruotsalo:
“The collection of individual brain signals is related to behavioral issues. Anyone who has gained this knowledge can gain a deeper understanding of a person’s preferences. We’ve already seen some trends. People buy “smart” watches and similar devices that can record heart rate and so on, but are we sure the data isn’t generated that provides information to private corporations that we don’t want to share? “
“I see this as an important aspect of academic work. Our research shows what’s possible, but we don’t do things just because they can. It’s an area that in my view needs to be addressed. regulate public guidelines and policies. If they are not adapted, private companies will continue to do so. ”
The scientific article on “Brain-Supervised Image Editing” was published June 19, 2022, at CVPR 2022 (Computer Vision and Pattern Recognition), the most renowned international conference in the field.
Brain-Guarded Image Editing
Article Publication Date
Disclaimer: AAAS and EurekAlert! is not responsible for the accuracy of news releases posted by EurekAlert! by contributing to institutions or for the use of any information through the EurekAlert system.