Neuroscientists claim that the biological principle of homeostasis will create intelligent robots, with feelings and thoughts.
In the modern world, robots experience no more feelings than a stone immersed in novocaine. However, there may be a way to teach robots to feel, say neuroscientists Kingson Man and Antonio Damasio. To do this, you just need to create a robot that senses danger and understands the threats to its own existence. Then he will only have to develop feelings that will determine his behavior, which is necessary to ensure survival. "Modern robots lack senses," Meng and Damasio write in their article in Nature Machine Intelligence. "They cannot determine the state of their internal processes in the mental space."
Scientists have proposed a way to endow machines (such as robots or humanoid androids) with "artificial equivalents of senses." In fact, robots should be forced to adhere to the biological principle of homeostasis - to instill in them the desire to survive. Awareness by artificial intelligence of its internal processes in the framework of survival is tantamount to a robotic version of feelings.
Kingson Man and Antonio Damasio are confident that senses will not only give robots a semblance of an instinct for self-preservation, but also allow artificial intelligence to more accurately mimic human intelligence.
This new intelligence will be suitable for a wide variety of tasks. Making robots truly smart is possible only through feelings, Meng and Damasio believe, and feelings are inherent only in those who seek to survive. When people keep the robot in working order (all its wires are connected, it receives enough energy, does not overheat and does not freeze), it does not need to worry about self-preservation. Therefore, he does not need feelings, but only calmly reports on faults.
Feelings motivate living beings to search for optimal conditions and conditions necessary for survival. Artificial intelligence, aware of its own vulnerability, should do the same.
The ability to create machines with feelings comes from recent developments in two key research areas: soft robotics and deep learning. Advances in soft robotics can bring feelings, and new deep learning techniques will enable the sophisticated calculations needed to translate those feelings into survival behaviors.
Deep learning is a modern descendant of the old idea of artificial neural networks - sets of interconnected computing elements that mimic the nerve cells of a living brain. Patterns in one layer are passed to the next level and then to the next, which allows the machine to recognize patterns in patterns. Deep learning enables robots to classify patterns into categories, identifying objects (such as cats) or determining whether a CT scan is detecting signs of cancer or other disease.
Promotional video:
By computationally representing the state of the environment, a deep learning machine could transform input data into a picture of the emerging situation. Such a smart machine, as Meng and Damasio point out, can “connect sensory modalities”: for example, recognize how lip movements (visual modality) correspond to vocal sounds (auditory modality).
However, the ability to feel your inner states is useless without a threat to existence. If the robot is made of soft materials with built-in sensors, it may become afraid of being cut or hurt and will engage in a program to avoid injury. Moreover, a robot capable of assessing existential risks can learn to develop new methods of protection instead of relying on existing programs of action. Developing new methods of self-defense can also lead to improved thinking skills.
Therefore, self-defense can motivate robots in the same way as Isaac Asimov's famous three laws of robotics: “A robot cannot harm a person or, by its inaction, allow a person to be harmed. A robot must obey all orders given by a human, except when these orders are contrary to the First Law. The robot must take care of its safety to the extent that it does not contradict the First or Second Laws."
Author: Kirill Panov