By Devanshi Saxena
After gaining a repertoire in the field of cyber-space and remote-sensing, artificial intelligence (AI) is extending its network to decode the intricacies of the inexplicable human brain. The cognitive neuroscientists from all over the world are specifically working with the techniques of artificial intelligence to analyse and understand the multifaceted functions of our intelligence system—the brain. At the 25th annual meeting of the Cognitive Neuroscience Society (CNS), Aude Oliva and her colleagues from the Massachusetts Institute of Technology (MIT) presented their work on human image recognition and the role played by contextual clues in the backdrop of this recognition mechanism. Artificial neurons were utilised along with neural network models to trace the pathway of the various elements that contribute in identifying a specific place or object as perceived by our senses.
The prime objective of the neuroscientific experiments
As elucidated by Olivia and her team, neuroscientists and computer scientists seek answers to similar fundamental questions. Both, the human brain and the computer have a similar internal configuration which is a complex system made of individual functional components. In case of computers, these components are called ‘units’ and in the case of the human neural system, ‘neurons’ are these structural and the functional components. Through the set of conducted experiments, Olivia and her team are aiming to discern the exact implications which are determined by these components. As a part of the experiment, more than ten million images were studied and an artificial network was trained to recognize and process approximately 350 different places including a kitchen, park, living room and a bedroom.
Major inferences drawn from the research
The experiments focused on providing extensive training to the artificial network to comprehend objects and places, for example, to understand that bed as an object is associated with a bedroom. An interesting discovery of the research was the fact that alongside the basic associations, the artificial network also learnt to recognize people and animals with the places, such as dogs in parks and cats in living rooms. When furnished with loads of data, the machine intelligence program can easily ascertain the refined levels of contextual learning. It is not possible in case of the human neural system to dissect the super-fine levels, but it is possible to achieve a similar feat with the computer models. Through a series of these experiments, cognitive neuroscientists have been able to accurately make conclusions about the functioning of the real human brain. The artificial neural networks as developed by the scientists functioned as ‘mini-brains’ and when these were studied, changed, evaluated and compared against the responses of the human neural system, it became possible to ascertain the intricacies of the nervous system.
Implications of the experimental results
A complicated scheme of nerves and ganglia, our nervous system is not a straightforward and easy-to-comprehend organ system. According to Nikolaus Kriegeskorte of Columbia University, who is chairing the symposium, “Neural network models are brain-inspired models that are now state-of-the-art in many artificial intelligence applications, such as computer vision.” Millions of signals originate from the retina, then pass through a sequence of neurons and are then interpreted by the central nervous system. The neural network models are a mode to get a generalised idea of how individuals perceive the world and its objects around them.
Future use of artificial network models
The neural network models can to an extent predict how the deep-seated neurons in the brain respond to sensory visions. The artificial networks cannot flawlessly imitate a human brain, but they are of immense assistance in developing a fundamental awareness regarding the scope and use of artificial intelligence.