Researchers at the Japan ATR International Electric Communication Advanced Technology Institute and Kyoto University in Japan said they created an AI that can read human brain waves. They built a neural network that can interpret and accurately reproduce images or imaginary images that one sees.
Researchers at the Japan ATR International Electric Communication Advanced Technology Institute and Kyoto University in Japan said they created an AI that can read human brain waves. They built a neural network that not only reads but also rebuilds your mind.
Specifically, according to the ZME Science website, "The research team created an algorithm that can interpret and accurately reproduce images or imaginary images that one sees."
A structural diagram of depth image reconstruction. The pixel values ​​of the input image are optimized such that the DNN features of the image are similar to those decoded in the fMRI activity. A depth generator network (DGN) can choose to combine with the DNN to produce a natural image, where the optimization is performed in the input space of the DGN. Credit: bioRxiv (2017). DOI: 10.1101/240317
The title of the research paper is "Deep image reconstruction from human brain activity," the paper writes, in which the researcher can copy an image based on the scene that a person is observing. The images created by these AIs are not exactly the same as what people actually see, but they vaguely express human thinking. However, AI can still use brain waves to reconstruct these images.
Although it may take decades to actually use it, this technology takes us one step closer to creating systems that can read and understand human thinking.
Trying to tame a computer to decode mental images is not a new idea. In fact, this research has been going on for many years, and researchers have been trying to reconstruct images of the brain since 2011, such as movie clips, photos, and even dreams. However, all previous systems were limited in scope and capabilities. Some can only deal with narrow areas such as face shapes, and some can only reconstruct images through pre-designed images or categories (for example, "birds", "cakes", "people", etc.). So far, all of these techniques require pre-stored data; they work by matching the brain activity of the subject to the brain activity of the previously recorded person while viewing the image.
But researchers say their new algorithm can generate new, recognizable images from scratch, and even generate shapes that are only imagined by the human brain.
It all starts with functional magnetic resonance imaging (fMRI), a technique for measuring blood flow in the brain and using its results to judge neural activity. The team scanned the visual processing area of ​​3 subjects at 2 mm resolution. The scan was performed several times. In each scan, the three subjects were asked to view more than 1,000 pictures, including a fish, an airplane, and some simple color graphics.
A new algorithm that uses brain activity to reconstruct (the bottom two rows) the observed image (first line). Image source: Kamitani Lab
The research team's goal is to understand the brain activity that responds to the image and ultimately generate images that produce similar responses in the brain through computer programs.
The team has recently begun to achieve results. The researchers used a deep neural network (DNN) and several layers of simple processing elements instead of presenting images to the subjects one by one until the computer got the correct results.
Yukiyasu Kamitani, the lead author of the study, said: "We believe that deep neural networks are a good representation of the brain's hierarchical processing."
“With DNN, we can extract information from different levels of the brain's visual system, from simple light contrast to more meaningful content, such as faces.â€
Using the decoder, the researchers created the brain's response to the images in the DNN. Then, they no longer need fMRI imaging measurements and no longer use DNN translation as a template.
Next is a repetitive process in which the system creates an image in an attempt to make the DNN react similarly to the desired template—whether it's an animal or a colored glass window. This is a trial and error process in which the program starts with a neutral image and slowly improves over 200 iterations. To understand how close it is to the desired image, the system compares the difference between the template and the response of the DNN to the generated image. Such calculations allow it to be improved pixel by pixel towards the ideal image.
To improve the accuracy of the final image, the team used a “deep generator network†(DGN), a pre-trained algorithm that creates realistic images from the original input. Essentially, DGN puts the finished details on the image to make it look more natural.
After the DGN completed the repair of the photo, a neutral human observer was asked to evaluate the work. He will be shown two pictures to choose from and asked which image is algorithmically reconstructed. The author writes in the paper that human observers can correctly select system-generated images 99% of the time.
The next step is to combine all the work with the process of “reading mindsâ€. They asked the three subjects to recall the images they showed before and scan their brains. The process is a bit tricky, but the results are still exciting – this method doesn't work well for photos, but for shapes, the generator creates recognizable images at 83% of the time.
It should be noted that this work looks very neat and cautious. Their systems work very well, and the bottleneck is not in software, but in our ability to measure brain activity. Perhaps we need to wait for the emergence of better fMRI imaging technology and other brain imaging technologies.
Description of Right Angle D-SUB Connector
The right angle d-sub is an ideal connector solution for applications where space is at a premium and there is little to no room for connectivity & cabling. The 90° angle cable exit allows for increased mounting options and versatility of design for those tighter, more compact application needs such as computer servers and industrial robotics.
Antenk RIight's Angle Standard D-SUB Connector Series Including:
Standard D-Sub Right Angle Machined
Standard D-Sub Right Angle Stamped
Standard & HD D-Subs SMT Right Angle & RA Zero Footprint Through Hole
VGA over Dual PS/2 ports Stamped Contacts
Antenk RIight Angle High Density D-SUB Connector Series Including:
High Density D-Sub Right Angle Stamped
High Density D-Sub Right Angle Machined
Dual-Port D-Sub
Standard & HD D-Subs SMT Right Angle & RA Zero Footprint Through Hole
High Density D-Sub Right Angle Stamped
Antenk's RIight Angle D-SUB Connector Options
Number of Rows
Shell Size
Mounting Style
Packaging
Gender
Shell
TARGET MARKET / APPLICATION
Communications
Base Stations
Switching
Transmission
Asymmetric Digital Subscriber Line (ADSL)
Data
Desktops/ Laptops
UPS, Storage systems
Routers, Servers
Printers, Copiers
Consumers
Consumer Electronics
Set-top-boxes
Energy meters
Industrial & Instrumentation
Robotics
Control Drives
Power Supplies
Medical Instruments
Test Equipments
POS & Handheld terminals
Renewable Energy
Surveillance Camera
Office Automation
Parking Meters
Gaming Machines
Military
Military
Avionics
Military Equipment
Standard Density right angle D Sub Connector machined contacts, Right Angle High Density D-SUB Connector,Standard Density right angle D Sub Connector stamped contacts
ShenZhen Antenk Electronics Co,Ltd , https://www.antenkconn.com