Another key challenge in the study of modern artificial intelligence is known as transplant learning. Artificial factors need the ability to build on existing knowledge in order to make sensible decisions in order to deal effectively with new situations. People are already good at this – a person who knows how to drive a car, use a laptop or lead a meeting is usually able to cope in the face of an unknown vehicle, operating system or social situation.
Researchers are now beginning to take the first steps to understand how this could be possible in artificial systems. For example, a new class of network architecture is known asprogressive network“Can use the information learned in one video game to learn another. The same architecture has also been shown to transfer data from the simulated robot part to the real-life arm, greatly reducing training time. Fascinatingly, these networks have some similarities models for learning sequential tasks in humans. These compelling links suggest that future artificial intelligence research has a good chance of learning about the work of neuroscience.
But this exchange of information cannot be a one-way street. Neuroscience can also benefit from artificial intelligence research. Take the idea of reinforcement learning – one of the key approaches in current artificial intelligence research. Although the original idea came from the theories of animal learning in psychology, it was developed and developed by researchers in machine learning. These later thoughts reverted to neuroscience to help us understand neurophysiological phenomena such as dopamine neuronal firing properties in a mammalian base.
This back and forth is necessary if both fields continue to build insights into each other and create a virtuous circle in which artificial intelligence researchers use neuroscience ideas to build new technology, and neuroscientists learn about the behavior of artificial substances to better interpret the biological brain. This cycle is likely to be accelerated by recent advances such as optogenetics, which allow us to accurately measure and manipulate brain function, resulting in a huge amount of information that can be analyzed with machine learning tools.
Therefore, we believe that distilling intelligence into algorithms and comparing them to the human brain is now vital. Not only could it strengthen our efforts to develop artificial intelligence, a desirable tool to create new knowledge and advance scientific discoveries, but can also allow us to better understand what is going on inside our own heads. It could shed light on some of the most enduring mysteries of neuroscience, such as the nature of creativity, dreams, and perhaps even one day, consciousness. With so much at stake, the need to combine neuroscience and artificial intelligence is now more urgent than ever.
Download the whole paper here.