When you think of Artificial Intelligence, the first thing that comes to mind is either Robots or Machines with Brains or Matrix or Terminator or Ex Machina or any of the other amazing concepts having machines that can think. This is an appropriate but vague understanding of Artificial Intelligence. In this article we’ll see what A.I. really is and how the definition has changed in the past.
While generalizing or defining a new concept/discipline, it is beneficial to set right goals and define the field by what we hope to achieve with it, along with how will we achieve it. Since Artificial Intelligence is a vast topic, embodying the knowledge of many domains of science, it’s definition has evolved rapidly in the recent years. The diagram below contains 8 definitions from different textbooks. Each of these definitions describe what is A.I. in a different sense. Lets understand how.
Credit — Stuart Russell and Peter Norvig
The various definition have been grouped into 4 different dimensions or schools of thought, based on what we want our A.I. to achieve and how we measure its success. As can be seen in the diagram, the definitions set the goal of Artificial Intelligence, either based on thought process and reasoning(top row), or based on behavior(bottom row). Further, the definitions are also grouped in terms of how we measure the success of Artificial Intelligence. This can be either in terms of human intelligence(left column), or against the idea of rationality or ideal concept of intelligence. The difference in human thinking and rationality is that, the former must be an empirical science, involving hypothesis and experimental confirmation and the latter involves a combination of mathematics and engineering.
Each of these 4 approaches, comprising Thinking and Acting, Humanly and Rationally, have been followed and each of them provides valuable insights into the field of Artificial Intelligence. We will now see what will systems, in each of these dimensions, look like and how feasible they are.
Systems that can think humanly
For following this approach, one first needs to understand how humans think. Knowing the internal working of human brain can be achieved by introspection or psychological experiments. This, in itself, is a vast interdisciplinary field, known as Cognitive Science. It brings together computer models from AI and experimental techniques from psychology to try to construct precise and testable theories of the workings of the human mind. Therefore, this definition is also known as Cognitive modelling approach. Today, Artificial Intelligence and Cognitive Science are two separate fields, but they continue to fertilize each other, majorly in the areas of vision, natural language, and learning.
Systems that can act humanly
This definition came into being when Alan Turing proposed the Turing Test. A system passes this test, if it can fool a human interrogator by depicting intelligent behavior. By intelligent behavior, we mean achieving human level performance in cognitive tasks.
Roughly, an A.I. system passes the Turing test, if during an interrogation, the human interrogator is unable to tell, whether it is interrogating a human or an A.I.
Such a system would require to have major components of A.I., including natural language processing, knowledge representation, automated reasoning, machine learning, robotics and computer vision. Seeing the underlying complications, no major effort has been made in trying to make such a machine.
Systems that can think rationally
The aim of this approach is to build upon programs that represent “right thinking”, to create intelligent systems. This “right thinking” or irrefutable reasoning processes, is defined in coding (in mathematical terms) using logic or laws of thought. Therefore, this approach is also known as Laws of Thought approach.
The basic problems that make this approach infeasible are:
- Not all knowledge can be expressed with logical notations (especially when knowledge is not 100% certain).
- It can lead to computational blow up, as without guidance, there are many reasoning steps that can be tried.
Systems that can act rationally
This approach involves creating systems that act in a way which maximizes its chances of achieving its goal, given the available information. These systems are known as Rational Agents, such that they perceive the environment and act so as to achieve the best outcome, or when there is uncertainty, the best expected outcome.
The study of AI as rational agent design therefore has following advantages.
- Unlike “laws of thought” approach, in which whole emphasis is on correct inferences, this approach achieves rationality by using correct inference as one of the mechanisms and not a necessary one. This property makes this approach more general.
- This approach is feasible for scientific development than approaches based on human behavior or human thought, as the standard of rationality is clearly defined and completely general. Human behavior, on the other hand, is well-adapted for one specific environment and is the product, in part, of a complicated and largely unknown evolutionary process that still may be far from achieving perfection.
The study of design of rational agents is relatively more general and feasible. However, all the approaches/directions to define Artificial Intelligence are useful in understanding its complexity and components, and also what it takes to truly create Artificial Intelligence.
Today, the state of the art, has reached a level , where we have A.I. that are world champions in Go, Chess, Checkers, etc, they can order food, book a cab, translate text, recognize people, play poker, and what not. But achieving the true potential of A.I. is still a long way to go.
If you liked this article, be sure to click ❤ below to recommend it and if you have any questions, leave a comment and I will do my best to answer.
I’ll soon be writing more on the study of designing rational agents and use of machine learning in Artificial Intelligence. So, for being more aware of the world of A.I., follow me. It’s the best way to find out when I write more articles like this.
You can also follow me on Twitter at @Prashant_1722, email me directly or find me on linkedin. I’d love to hear from you.
That’s all folks, Have a nice day :)
Credit
Content for this article is inspired and taken from, Artificial Intelligence, A Modern Approach. Stuart Russell and Peter Norvig. Third Edition. Pearson Education.