Artificial Intelligence Modern Approach – 001
This series of posts corresponds to study through a most famous book of the exciting field of artificial intelligence. Some might say, its a Bible for AI. I will be updating my takeaways from this book in this series
Intelligence is considered as most exciting and distinguishing feature of any living organism. It separates us from nonliving things and allows us to manipulate the exciting place we call Home, The Mother Earth. But what can be considered as intelligent? Can we define intelligence? Can we create it?
Intelligence is nothing but Perceiving, Understanding, Predicting and manipulating the surrounding environment to our advantage. Now, by definition, any living organism can be considered as Intelligence. But with technological advantage, can we create such thing in this world? Artificial Intelligence is a branch of science which tries not only to reproduce the performance of intelligence task but also tries to build entities which can do such things.AI agents are considered rational agents. A rational agent is said to be rational when it performs right thing, given system knows what is the right thing. It is an ideal situation where the system will always do an action which according to it is right. Since the action of the system depends on facts provided to it and protocols decided external to the system, system’s actions are based on logical reasoning.
Scientists are pursuing AI in 4 different ways:
- Acting Humanly: AI is considered as a field such systems are made which performs tasks where intelligence is required. Such machines can be said to be succeeded in Turing test. Such machines must have capabilities like natural language processing for communication, knowledge representation to store information, automated reasoning to draw conclusions and machine learning to find a pattern. Also if we consider total Turing test, then it must have computer vision for object recognition and robotics for object manipulation.
- Thinking Humanly: AI is the field where one is creating machines with minds which can automate activities associated with human thinking like decision making, problem-solving, learning etc..To make a system to think like humans, it is necessary to understand how the brain actually works. The combination of computer models of AI and experimental techniques from psychology to construct theories of human mind is the field of cognitive science. Even though cognitive science and AI are considered as two separate fields, they closely support each other like in the field of computer vision.
- Thinking Rationally: AI is approached as a study of making computations, perceive, reason and act. This approach tries to develop a system which can solve any solvable problem described in logical notations. Thus creating intelligent systems which can solve the problem or keeps trying going into forever loop. But its quite difficult to make every problem’s description formal. Also as parameters/facts of problem increases, it became computationally difficult in practice.
- Acting Rationally: Another approach for AI can be taken as a study of designing intelligent agents i.e. study of intelligent behavior in artifacts. The rational agent approach tries to get the best outcome if possible or else best-expected outcome. Thus it emphasizes on correct conclusion. The correct conclusion can be logically decided or can an action which was done because something has to be done just like reflex actions. Turing test allows an agent to act rationally. This approach is advantageous to others as the correct inference is a way of achieving rationality and it is responsive to scientific development.
Applications of AI can be found in various fields such as
- Robotics Vehicles
- Speech Recognition
- Autonomous planning and Scheduling
- Game playing
- Spam Filtering
- Logistic planning
- Machine Translation