ItR Reading Response: Artificial Intelligence for Robotics
What’s AI?
What’s the difference between AI and traditional algorithms?
The uniqueness of artificial intelligence is that it enables the robot to think and make decisions by itself. The input and output are not fixed code that has been written by a human. We don’t need AI in a structured environment. In this kind of environment, all the things are located as planned. It’s simpler and more efficient to give an accurate location and description of the object to the machine but not learn the feature by itself. In contrast, AI needs to be adopted in an unstructured environment. In this environment, all the objects are not assigned in advance, but randomly appear. It asks the machine to have the ability to recognize the object and the environment. In the real world, most of the situations that we are facing are unstructured environments. For instance, a messy living room, a theme park, or a restaurant in the shopping mall. If we want to adopt robots around this place, changing the code one by one is inefficient. It’s better to let the robot can adapt to the environment by itself. I think this is also the reason why AI has gone through a burst of development after the miniaturization revolution and performance growth in computers.
How to AI?
The OODA Model
The OODA model comes from the US air force. It also has a great contribution to robotic desition making. The initial content of the OODA is the process of Observe- Orient-Decide- Act loop. When we adapt it to the robots, the observation becomes receiving information from the sensors. The orientation becomes processing the data using a certain model or algorithm. The decision here means the outcome of the algorithm and the Act means giving the order to the servo and motor to move.
Subsumption Model
The subsumption model is similar to the human body. We need to ensure the basic organ work properly before considering how to finish the assignment or write an essay. When comes to robots, we need to divide what kind of activity is the priority that we need to consider about. The first thing is to make sure the robot will not “die“ because of falling down the stairs or hitting an object at an extremely high speed. Then comes the basic functions of the robots like cleaning the floor or following the given lines. The last consideration is the interactive part of the entertainment part which is how to let human knows better what the robot is doing. For instance, playing some music or displaying the UI on the screen.
Ethics and Risk of AI
From Before to After
Many people worry about the future of AI. They are considering after AI gets self-conscious, will they rule humans? By viewing the historical development and current situation of AI we can see that the basic of AI is a self-generated algorithm. It can identify the things that are not described in advance and find the way to do the task. According to the needs pyramid, the AI robot won’t have the desire for wealth and power on its own. The only thing that it ought to do is complete the task. The difference is that the robot with AI can adapt to more environments (unstructured environments). If the day that robots kill all humans finally comes, it’s not the fault of the AI, but the code written by humans behind it.