Ticker

6/recent/ticker-posts

Four categories of AI

Four categories of AI

AI falls into four categories:

 i.                    Thinking humanly

ii.                  Thinking rationally

iii.                 Acting humanly

iv.                Acting rationally

      

    i.            Thinking Humanly:

            This means trying to make computers think like humans do—so they can have their own kind of "mind."

Here are a few quotes that explain AI:

    Haugeland (1985): He said AI is the exciting effort to make computers that can think, just like humans do.

     Bellman (1978): He described AI as automating activities like decision-making, problem-solving, and learning, which are things humans do when they think.

        ii.             Thinking Rationally:

            This is about teaching computers to make the best possible decisions, like how humans try to do things in the smartest way.

 Charniak and McDermott (1985): They said AI is about studying the mind and using computer models to understand how our brains work.

  Winston (1992): He said AI is about studying the steps or calculations that help machines see, think, and act.

iii           Acting Humanly:

          This means trying to make machines do things that normally require human intelligence, like   understanding language or recognizing faces.

    Kurzweil (1990): He said AI is about creating machines that can do things that need human intelligence, like solving problems or thinking logically.

    Rich and Knight (1991): They said AI is focused on making computers do things that humans are better at right now, like understanding speech or driving a car.

      iii.            Acting Rationally:

               This is about teaching machines to always make the best possible decisions, even if they don't   think exactly like humans do.

    Poole et al. (1998): They said AI is the study of designing smart systems or "intelligent agents" that can solve problems.

    Nilsson (1998): He said AI is about creating machines or "artifacts" that can show intelligent behavior, like thinking or learning.

Acting Humanly and the Turing Test:

The Turing Test (1950):
Alan Turing asked, "Can machines think?" or "Can machines behave intelligently like humans?"

He created a way to test this called the Imitation Game (now known as the Turing Test). In this test:

·       A human asks questions through writing, and the computer responds in writing.

·       If the human cannot tell whether the answers are coming from a person or a machine, the machine passes the test.

For a computer to pass the Turing Test, it needs several abilities:

1.      Natural Language Processing: The computer must understand and communicate in human language, like English.

2.      Knowledge Representation: The computer must be able to store and remember information it knows or hears.

3.      Automated Reasoning: The computer must use the information it knows to answer questions and make new conclusions.

4.      Machine Learning: The computer must be able to learn from new experiences and find patterns in data.

To pass the Total Turing Test (which is even harder), the computer would also need:

·       Computer Vision: The ability to see and understand objects.

·       Robotics: The ability to move and manipulate objects, like a robot.

In short, the Turing Test checks if a machine can think and behave like a human, based on how well it understands language, learns, and makes decisions.

Thinking Humanly and the Cognitive Modeling Approach:

Thinking Humanly (Cognitive Modeling):
To say that a program thinks like a human, we first need to understand how humans think. We need to study the human mind closely.

There are three ways to do this:

1.      Introspection: Trying to pay attention to and understand our own thoughts as they happen.

2.      Psychological Experiments: Watching people in action to understand how they think and solve problems.

3.      Brain Imaging: Using technology to see how the brain works while it is thinking.

The goal is to compare how a computer solves problems with how humans do the same thing. By doing this, we can see if the computer is thinking in a way similar to humans.

Cognitive Science is a field that combines computer models from AI and techniques from psychology. It tries to build and test theories about how the human mind works.

·       AI and Cognitive Science work together, especially in areas like understanding vision (seeing things) and language (understanding and speaking).

Once we have a clear theory of how the human mind works, we can write it as a computer program. If the program behaves the same way a human would (based on how we observe humans think), it gives us evidence that the program is working like a human brain.

For example, Allen Newell and Herbert Simon created a program called the "General Problem Solver" (GPS), which was designed to solve problems the same way humans do.

In short, this approach is about studying the human mind and creating computer programs that can think the same way humans do.

Thinking Rationally and the “Laws of Thought” Approach:

Thinking Rationally:
This approach is about teaching machines to think in a logical and reasonable way, just like humans try to do when they use correct reasoning.

One of the earliest thinkers to talk about "right thinking" was Aristotle. He created rules for logical reasoning, which are still used today. He came up with things like syllogisms, which are simple patterns of arguments that always lead to the right conclusions if the starting information is correct.

For example, the classic syllogism goes like this:

·       Socrates is a man.

·       All men are mortal.

·       Therefore, Socrates is mortal.

This is a basic form of logic, where if the first two statements are true, the third must be true too.

However, there are two big problems with this approach:

1.      Turning informal knowledge into formal logic: It's not always easy to take everyday knowledge (which isn’t always 100% certain) and put it into a strict, formal system of logic.

2.      The difference between theory and practice: Even if a problem can be solved using logic in theory, it’s much harder to solve it in real life because of the complexity of real-world situations.

In short, this approach is about trying to make machines think logically and reasonably, but it's not always easy because real-life knowledge is often uncertain or too complex for simple rules.

Acting Rationally and the Rational Agent Approach:

Acting Rationally:
An agent is anything that acts. All computer programs do something, but a computer agent has higher expectations—it needs to do more than just process tasks. It should be able to:

·       Act on its own (autonomy).

·       See and understand its surroundings (perception).

·       Keep going over time (persistence).

·       Adapt to changes around it.

·       Set and work toward goals.

A rational agent is one that acts in a way that gets the best possible outcome. If the outcome is uncertain, it tries to get the  expected result.

For example, in the “laws of thought” approach, the focus was on making sure inferences (conclusions) were always correct. But sometimes in life, there isn’t one clearly correct choice, and a decision still needs to be made.

For instance, if you touch a hot stove, you quickly pull your hand away. This is a reflex action that is better and faster than thinking carefully about what to do.

What does it mean to "behave rationally"?

To behave rationally means to choose the best action to reach your goal based on what you know or believe. Here’s an example:

·       Goal: I don’t want to get wet in the rain.

·       Action: I bring an umbrella.

Is this rational? It depends on what I know or believe.

·       If I know the forecast says it will rain and I believe it, then bringing an umbrella is a rational choice.

·       But if I haven’t heard the forecast and don’t think it’s going to rain, then bringing an umbrella would not be rational.

Even if I act rationally, it doesn't always guarantee success. Here’s an example:

·       My goals: (1) Don’t get wet if it rains; (2) Don’t look silly by bringing an umbrella when it’s not raining.

·       My belief: I believe it will rain because I heard the forecast.

If it rains, my rational behavior (bringing the umbrella) helps me achieve both goals. But if it doesn’t rain, my action doesn’t help with goal #2 (not looking silly).

So, the success of behaving rationally depends on what I know and believe. If I don’t have enough or correct information, my rational choice might not lead to the best result.

Post a Comment

0 Comments