According to reports, recently, a team of scientists from Columbia University conducted a very bold experiment - a new transformation of robots and artificial intelligence . Different from the previous design of a program for the robot and the surrounding environment, this time Researchers enhance self-awareness in robots and artificial intelligence
What does this mean? To put it simply, scientists decided to let the robots think for themselves, let the robots decide for themselves what their hands and feet can do, and at the same time observe how the robots will react when the external environment or events change. Come on, I have to say, this is really crazy.
Scientists build self-awareness for robots. How bold is this?
researchers used the robot's visual system to add autonomous programs to perform self-learning based on the robot's external form. During this learning process, the robot will continue to perform motion modeling, and the robot will learn to walk, run and even various motion planning. You can also learn to use your hands for a variety of complex tasks. What's even more powerful about
is that the robot can deal with various extremely complex situations through the built kinematics model, such as how to avoid danger and how to avoid external attacks. This is just like the human brain. When humans encounter danger, they will If I instinctively do what is best for me, then the robot will think about what I can do.
A journal about robots, "Science Robotics", has published a paper on robots' self-thinking and self-learning. This article describes that robots can do themselves without human help (programs to perform specific tasks). Scan and build a body model. On this basis,
simulates various motion models. In the end, the robot can find the optimal solution route like humans. In this process, it can also realize self-correction and self-correction. Through continuous learning and accumulation, the final robot can be completed. task.
So, how does a robot with "autonomous consciousness" learn?
In the Creative Machine Laboratory of Columbia University, scientists have conducted relevant experiments. The researchers placed a robot in a room surrounded by mirrors. This robot has an autonomous neural network and self-learning ability. At the beginning At that time, the robot was like a baby, very curious about the outside world. Through the reflection of the mirror, it realized that it was itself in the mirror.
As the robot continued to observe itself, it began to gradually make various actions and explore the connection between the program and the actions. After learning for 3 hours, the robot had a full understanding of itself.
For example, it knows what it looks like, how tall it is, what kind of movements it can make, how to pass through doors smaller than itself, and other issues. And as this self-understanding continues to deepen, the accuracy and The capabilities will be improved accordingly, so that the robot can complete various complex tasks.
In fact, humans invented robots to complete complex and dangerous tasks instead of humans. Current robots rely more on human operation and judgment to work. Only in specific working environments can they be fully automated.
If a robot can recognize and learn by itself in an unfamiliar environment, no matter what emergencies occur, the robot will not require manual intervention. For example, the robot will replace humans in the factory to inspect various problems and faults. It can also Make the most correct judgment in an extremely complex environment.
So, can robots generate intelligence?
In many film and television works, we can often see that robots generate intelligence and eventually turn against customers. So can robots in reality really reach this level?
The answer is no, at least the current answer is no.Even the robot mentioned above is just a program, which is fundamentally different from a robot that can produce human intelligence. Generally speaking, a robot with self-observation and self-learning can only focus on a certain aspect. It has no self-awareness. And emotional, the process of self-learning is only aimed at the construction and accumulation of motor behavior patterns.
It also needs humans to give it specific task goals, but this robot does not need humans to teach it how to do it. You only need to say: "Bring me the drink on the table", and the robot will learn how to use both hands to hold things. , and then complete the action of getting the drink, further improving the efficiency of the robot. The robot neural network only simulates the way humans think, and ultimately turns this way of thinking into a working mode of the robot.
Judging from the current level of human technology, it is almost impossible to make robots have emotions and thinking, so we should not worry too much. After all, robots and AI ultimately serve humans and help humans improve production and life. Efficient. So, do you think that as scientists continue to upgrade the "new functions" of robots, does it pose a threat to humans? Welcome to tell us your opinions!
