People and computers see three-dimensional objects the same way

(ORDO NEWS) — Researchers have found similarities in the way artificial intelligence algorithms and the human brain perceive visual information about three-dimensional objects.

Researchers are striving to ensure that artificial intelligence is similar to natural in some aspects. Now, artificial neural networks have found signal patterns that are eerily similar to what happens in the brain.

Scientists have long been trying to teach artificial intelligence to see in the same way as humans. With the help of deep neural networks, researchers have already made significant advances in object recognition. This requires the use of high performance graphics processors (GPUs) and large training samples of videos and images.

The authors of the new work decided to compare to what extent the principles of processing visual information are similar in artificial and natural neural networks. Scientists examined neurons in the V4 visual field, the first region of the brain that processes visual information. Then the authors took a close look at the image processing processes in the AlexNet neural network and found that the signal patterns in the artificial and natural neural networks are “eerily” similar to each other.

According to the researchers, AlexNet and similar deep networks were developed on the basis of multi-level visual information processing systems in the brain. The close similarities that the researchers observed between artificial neural networks and the brain may indicate the possibility of using this connection to develop more advanced algorithms and better understand how the brain works.

“Artificial neural networks are the most promising modern models for understanding the processes taking place in our brain. Conversely, the brain is the best source of ideas by which existing artificial intelligence can approach artificial intelligence in its capabilities, ”says Ed Connor, professor of neuroscience at Johns Hopkins University.


Contact us: [email protected]

Our Standards, Terms of Use: Standard Terms And Conditions.