Numerous studies have been published about Artificial Intelligence (AI) during the latest years. However, what relationship does the human brain have with the AI? We can say that Deep Learning is based on the mental processes of human beings and, moreover, those processes are investigated in the field of neuroscience and/or cognitive sciences. Therefore, when we talk about Deep Learning, we talk somehow about the functioning of the brain.
Inspiration for evolution?
Even if it is not essential to know how the biological brain works, this will help us to improve the architecture of the processes that specifically support the artificial neural networks. Although it is true that the artificial neural networks can get a real similarity with the human brain activity, the evolution of the artificial neural networks during the last years has been more related to a mathematical, statistical and algorithmic perspective than to neuroscientific studies.
In this sense, neuroscience deals with the more practical studies, that means, knowing the structures and functions at brain level, even if we don’t know for sure how the brain works. Some techniques such as functional magnetic resonance imaging, allows us to know the brain activity, how we react to certain stimuli, and even showing the areas of the brain which can be damaged. The electroencephalography also allows us to know the bioelectric activity of the brain almost in real time. Although these techniques are in the group of the most precise and commonly used, nowadays the signals are being improved thanks to AI. However, more tools are necessary to read the brain activity and are more focused on the neural activity of the brain.
Will it be possible to reach that reasoning capability in Deep Learning?
As we already know, the human being communicates through the language and this implies reasoning skills. For that reason, the research which complements neuroscience with Deep Learning may be crucial. Both fields together could obtain more reasoning capability and improvements in the processing part of the AI. It will be a positive feedback for both. In fact, when we work with Deep Learning, we can differentiate two types of learning. On one hand, that deep learning, where there are layers of artificial neurons and nodes. On the other hand, reinforcement learning, which is basically trial and error, and it is close to the concept of classic conditioning in psychology.
Some experts say that all we know about AI so far is based on what we knew about brain functioning 40 years ago. From the point of view of the technology, this is exciting. This means that it is possible to progress from a more neuroscientific field, and we know that there are upcoming years for researching.
Perhaps the exchange of ideas and the integration of the two techniques, both the neuroscience and the AI can lead us to find how we understand, think and remember, in a more scientific way. The day it is possible to understand how the brain works and more specifically the biological neural networks, perhaps it will be possible to reach the reflection and comprehension capabilities using artificial neural networks.