FALL, 2000



Jia Min Huang
ESL 91
Fall, 2000

Essay Topic: Research recent developments in the field of Artificial Intelligence, particularly concerning the COG project at the Massachusetts Institute of Technology (MIT). What is the purpose of the COG project? How does the COG project fit into the field of human-computer interaction?


Artificial Intelligence, AI for short, is the focus of present computer technology. The purpose of AI is to understand the nature of intelligence and to create systems, which can mimic human thought, understand speech, beat the best human chess player and other feats (An Introduction to Artificial Intelligence, nd). The goal of AI research is to make computers able to reason and solve problems in the same way humans do. Artificial Intelligence is a combination of computer science, physiology, and philosophy. The research into the areas of learning, of language, and of sensory perception has aided scientists in building intelligent machines.

AI has come a long way from its early roots. It began to intrigue researchers with the invention of the computer in 1943 (An Introduction to Artificial Intelligence, nd). After 50 years researching, scientists have got great success in exploring the area of AI. From machine vision to expert systems, there are a great deal of successful achievements, such as all kinds of intelligent robots, the research on Natural Language Processing and neural networks.

What is Intelligence?

The goal of Artificial Intelligence is to create intelligent computers. In that case, we have to understand the definition of intelligence before creating programs to make machines "smart". Human intelligence includes four central aspects: developmental organization, social interaction, embodiment and physical interaction, and multimodal integration (Our Research Methodology, nd).

  1. Development forms the framework by which humans successfully acquire increasingly more complex skills and competencies.
  2. Social interaction allows humans to exploit other humans for assistance, teaching, and knowledge.
  3. Embodiment and physical coupling allow humans to use the world itself as a tool for organizing and manipulating knowledge.
  4. Integration allows humans to maximize the efficacy and accuracy of complementary sensory and motor systems.

How to Make Computers Behave Intelligently?

Computers are not like human beings that are born with essential abilities for intelligence. To program a computer so that it exhibits intelligent behavior, we have to "determine the process used by humans to produce a particular type of intelligent behavior and then simulate that process on a computer" (Kasper, 1998, P.57.). This is a very difficult and huge job. Thousands of researchers and engineers as well as advanced technology are needed.

Neural Networks

The human brain is made up of a web of billions of cells called neurons, and understanding the brain's complexities is the aim of AI researchers. Although much of the working of the brain remains unknown, the complex network of neurons is what gives humans intelligent characteristics. Based on experiments with neurons, two researchers, Warren McCulloch and Walter Pitts developed a network theory. Using this theory, they designed electronic replicas of computerized neural networks. Computerized neural networks are good at solving problems which humans are good at solving. However, computers still are weak at complex problems or solving problems with no defined solution. Though, according to McCulloch and Pitts, neural networks may be able to learn and recognize patterns in the future (Neural Network, nd).

Artificial Intelligence Computer Applications

Currently, the most popular computer application of AI is Expert Systems. An expert system, as its name, is a system like a human expert that is able to do the work of a professional. Some fields such as forecasting weather and finding bugs in computer software have used computer experts instead of human experts.

A computer system has many merits. It can be trained quickly, has virtually no operating cost, never forgets what it learns, never calls in sick, retires, or goes on vacation. Moreover, an intelligent computer can consider a large amount of information that may not be considered by humans (The Scope of Expert Systems, nd).


Boxer is the first example of a "computational medium" for real people - not just for computer experts. It is based on a literacy model, created by DiSessa, which makes the computational medium useful to everyone. People learn it once, and use it forever. Boxer contains a complete but integrated set of facilities for the broadest possible range of human intellectual activities. These facilities include text and hypertext processing, dynamic and interactive graphics video, personal data management (including e-mail and net working), and programming (Boxer, nd).

Why Build a Human-like Robot?

A robot is an electro-mechanical device that can be programmed to perform manual tasks. Researchers built a robot with a humanoid form for two sets of reasons. For one reason, if we are to build a robot with human like intelligence, then it's necessary to make a human like body for it to be able to develop similar sorts of representations. For a second reason, if the robot has humanoid form, it will be both easy and natural for humans to interact with it in human like way.

Cognitive Systems

Cog is a single hardware platform which seeks to bring together each of the many sub-fields of Artificial Intelligence into one unified, coherent, functional whole (Overview of the Cog Project, nd). It is a new type of robot designed to behave in the same way as human child. Cog has an iron head and two iron arms connected to the body. And there is a set of microphones and video cameras in the head for hearing and sight. Cog's brain is made up of a network of microprocessors. All the processors work together to make Cog function properly. Cog is now still being created at the Artificial Intelligence Lab at the Massachusetts Institute of Technology.

Natural Language Processing

People traditionally communicate with computers by using programming languages or a set of commands. The appearance of natural language processing (NLP) allows people and computers to communicate in a natural, human language, such as English or Chinese.

There are two sub-fields of natural language processing research: natural language understanding and natural language generation. "Natural language understanding studies methods of allowing the computer to understand instructions in ordinary English and thereby to understand people more easily. Natural language generation involves efforts to have computers produce ordinary English so that people can understand them more easily"(Kasper, 1998, P.58-59).

My Response

Since childhood, I have seen many real or cartoon movies about robots with Artificial Intelligence. The outward appearance of some of those robots looked like human beings, having a head, two arms and two legs connected to the body. The robots also behaved like humans that can talk with humans, think, learn and do all kinds of work. When I was very young, I thought the robots essentially existed in the world. But, mother told me that, "those robots are only imaged, they are impossible to create". I believed her.

After doing this research paper about AI, I developed a lot of knowledge about today's development of computer technology. The first thing I realized is what mother said about robots is not absolutely true. Robots are possible to create. And scientists have done research to make robots for many years. Now, functional robots such as Cog and Kismet are almost finished.

AI is a broad topic. It consists of many different fields. Besides intelligent robots, scientists are also exploring to make Expert Systems, neural networks, Natural Language Processing and so on. In conclusion, the element that the fields of AI have in common is the creation of machines that can "think".

Scientists are now still on their ways of inventing intelligent computer. I believe they will get success in not a long time. Then we will use computers more conveniently. For example, we won't have to learn all kinds of complicated programming languages to communicate with our computers, but just "tell" the computers to do what we want. This would be very helpful for the Computer Science people. Now learning programming languages is a primitive part in Computer Science. Many people deem Computer Science as the hardest major in college, because the programming languages are difficult to learn. Moreover, we can make friends with computers. Then, whenever we have something weighing in our minds, but we feel shy to talk to parents or friends, we can share it with computers. This would be very amazing.


AI research has made a huge contribution to computer science. All the theories and insights of AI research have enhanced the development of computers and they will continue to set the trend for the future of computing. The products available today are only bits and pieces of what are soon to follow, but they are a movement towards the future of artificial intelligence. The advancements in the quest for artificial intelligence have, and will continue to affect our jobs, our education, and our lives.

Works Cited

Kasper, Loretta F. (1998). Interdisciplinary English. New York: McGraw-Hill, pp. 55-59.

Our Research Methodology (1999) [online]. Available:

Overview of the Cog Project (1999). [online]. Available:

Return to the Top of this Page

Mansoor Hayat
ESL 91
Fall, 2000

Essay Topic: Research recent developments in the field of Artificial Intelligence, particularly concerning the COG project at the Massachusetts Institute of Technology (MIT). What is the purpose of the COG project? How does the COG project fit into the field of human-computer interaction?


What is Artificial Intelligence? Artificial Intelligence is a study in the area of computer science, in which researchers are trying to create machines that could behave more like humans. Their attempt is to make computers do things that a human can do more easily and perfectly. Their goal is to make computers think, read, understand, recognize voices and be able to work according to human commands. The element that all AI researchers have in common is the creation of machines that are intelligent. In order to classify machines as "Intelligent", it is necessary for researchers to define what Intelligence is. Is Intelligence the ability to read, write, understand make common sense, generate new ideas and solve complex problems? Or is it something else? The best way to gauge the intelligence of a machine was stated by Alan Turing, a British computer scientist. Turing said, "A computer would deserve to be called intelligent if it could deceive a human into believing that it was human" ( 2705/history.html, 1997).

The term artificial intelligence was first introduced in 1956 at the Dartmouth conference. In 1957 the first version of a new program "The General Problem Solver" (GPS) was tested. The program was developed to solve common sense problems, which a human can solve easily (The history of AI). This was one of the steps towards the artificial intelligence goal of creating machines. Today with the invention of computers and 50 years of research into AI programming techniques, the dream of creating smart machines is becoming a reality.

Machines like Cog and Kismet are being worked on in the MIT AI lab. Cog and Kismet are basically robots with a head and two arms connected to a body. These projects are working so efficiently that they are able to mimic human thoughts, understand speech, recognize voices, and beat the best human chess players and so on. Cog and Kismet will continue to affect our jobs, play effective role in our education, and could be a source for us to learn different languages. Projects like Cog and Kismet are not finished yet but will be accomplished soon in the near future.

Why build a human-like robot?

The goal of artificial intelligence is to create machines that will be able to interact with people the way humans do. In one way or the other, researchers are trying to make a human like machine that has the capability of understanding and making common sense out of no where. AI research of 50 years has allowed the researchers to create robots named COG and KISMET.


The researchers of the MIT AI lab have been developing a robotic torso, called Cog. Their goal is to create a robot, which is capable of interacting with both objects and people in a human like way. Cog is designed to be intelligent, perceptive, able to judge and have motor control. The researchers came up with the thought that if they are to build a human-like machine, then it must have a human like body in order to develop human ways and to represent human feelings, understanding, and thoughts. The researchers are focusing on building a robot that is as human as possible. Cog is a two-armed machine with a head and over twenty actuated joints. The arms are completely symmetric, and they have no preference for one or the other. Cog has every part a human body should have, except legs but Cog is not perfectly human-like. Cog is not designed to feel emotions, regret; it is simply a machine that has some abilities that humans have.

The present prototype Cog was built in different stages. According to researchers there actually were no plans made, there were no blue prints of their research or drawings of the machine, Cog. Cog does not have a face, and the research goal is to design a humanoid face for Cog. With a well-designed face, Cog would be able to convey appropriate messages and interact with humans more meaningfully.

Moreover, Cog doesn't have a long-term memory. The robot cannot remember things that were done even two minutes ago. However, Cog is still being worked on and researchers will surely come up with something related to the machine's memory. The researchers are changing and improving the technology of the machines (


The sociable machine project at MIT is developing an expressive anthropo- morphic robot called Kismet. Kismet is designed to engage with people in natural and expressive face-to-face interaction (Cynthia). It's designed to express emotions, facial expressions, body posture, gaze direction and to have an infant's voice. Researchers have tried to create a baby. Kismet is equipped with visual, auditory and proprioceptive sensory inputs. Kismet is more human-like than Cog.

The machine Kismet is able to read, understand and react positively to the inputs given by people and/or the researchers. The machine is made of cameras, a microphone, and an articulatory synthesizer. Kismet is equipped with almost everything that could make Kismet perform more like a human. Kismet has been designed to support several social cues and skills that could ultimately play an important role in socially situated learning with a human instructor. Moreover, the approach of creating Kismet was inspired by the way an infant learns to communicate and start socializing with people.

Kismet is designed to be an altricial system, similar in spirit to a human spirit (Cynthia). That is, the robot starts off in a rather helpless and primitive condition and requires the help of a sophisticated caretaker to learn and develop. The relation between the caretaker and robot is completely social, much like the way in which a mother interacts with her infants.

Moreover, the primary and important issue for the researchers is how to establish a natural, real-time dynamic of interaction between learner and caretaker. From a broader perspective, this research not only aims at creating a learning machine, but also to build a system that a human can interact with and train in a natural and instinctive manner. (Sociable machines-Ongoing Research)

My Response:

The first time when our professor gave us the passage "Artificial intelligence" to read, my first reaction to the title was "Is this passage about the FBI, any thing related with investigation". But as I started reading the passage and then finished reading all of it, I ended up knowing something that I had never thought of and had never heard of.

It is not my aim to surprise you, but the simplest way I can summarize it is to say that there are now in the world machines that can think, that can learn and that can create. Overall, their ability to do things is going to increase rapidly in the visible future.

My personal knowledge about robots was that there are researchers who are trying to create robots that could do a lot of things like manufacturing cars, performing operations, and being in space projects. But robots that could actually talk and interact with people, express ideas, emotions, understand us was far from my knowledge. Reading the text for this project led me to do research on what AI is and how far they have gotten in their research of creating humanoid robots. I learned a whole lot about Cog and Kismet, the projects of the MIT AI lab. The researchers there are improving the technology of robots day by day.

While looking for information connected to my research project, I learned how to search the Internet. The Internet nowadays is the best source of information. A person looking for any information could just go on the net and get a whole lot of information for their search. I learned how to look for the different and the correct information. I learned a lot about what the MIT AI lab researchers are up to and how far they are in their research.

The future we will be facing will surely be different, unbelievable and hard to accept, but that will be reality. Everything will be mechanical; robots will be talking like humans, dealing with daily life aspects. Almost everything will be controlled by machines and computerized. For instance, just take a house totally controlled by a computer or a machine. You could just say things like "lights", "heat on/off", or "air-condition on/off" and there it will be done. A person practically won't have to do anything himself.

The major role in this technology will be voice recognition. Machines will have to recognize voices and be able to understand the voices of different persons. There are now software, programs, which are related with voice recognition. Some work well, but some are not useful. Sometimes they have difficulty in recognizing pronunciation or there is some thing else wrong.

The advancement in AI will continue to work on these projects. AI will affect our jobs, our education and our lives. We will be turning ourselves into machines. This was my first research project and doing this project I really learned a lot, gained my knowledge and learned about what's coming up in the future of technology and the way in which our lives will be changing in many different ways.


AI has always been on the pioneering end of computer science. Advanced level computer languages, as well as computer interfaces and word processors owe their existence to the research into artificial intelligence. The theory and insights brought by AI will set the trend in the future of computing. The products available today are only bits and pieces of what are soon to follow, but they are a movement towards the future of Artificial Intelligence.

The future of technology, the future of our lives, everything will be affected by what is created today. Machines, computers will be so intelligent that they will be able to deceive humans. They will play a role in every aspect of our lives. We can see our future just by looking at what the MIT AI lab researchers have already created, machines like Cog and Kismet.


Anika. Cog,

Cynthia. Kismet,

Cynthia (Sociable machines-Ongoing Research), view/html

Turing, Alan. An Introduction to Artificial Intelligence, 1997.

Return to the Top of this Page

Chun Chung Cheung
Fall, 2000

Project Topic: Research the development of Artificial Intelligence and how scientists have applied AI in the COG project at the MIT AI Lab. What is the function of the COG Robot? What is the goal of COG project and how COG Robot interacts with human being?


Because computer experts want to make computers work more intelligently to solve more complicated problems and perform tasks as the human brain does, they are developing and researching Artificial Intelligence (AI) continuously. Artificial intelligence technology is used to make machines respond to what they see and what they hear after executing the specific programming and advanced system. Moreover, AI Computers can learn and accumulate the knowledge from human beings and then they can simulate the human response in different situations. Actually AI researchers make machines that study like a baby. When computers store a vast database which is collected from interacting with people, they can be smarter and more intelligent as well as more human (AI on the Web, 2000).

This research paper will focus on the development of AI, the definition of human intelligence, Neural Networks, Artificial intelligence research, expert systems and will explain how computer experts are using Artificial Intelligence technology to make computers perform human-like behavior and interact with people. Moreover, it will describe the MIT AI research project on COG and Kismet, which studies robots and how they simulate human behavior.

The History of Artificial Intelligence

In 1956 John McCathy became the father of AI. He invented the theory of AI and organized a “Dartmouth summer research conference held in August 1955” (The History of Artificial Intelligence, N.d.) to attract the computer experts and other people who were interested in machine intelligence. The conference invited these people for a month of brainstorming. The conference goal involved the study of a wide range of topics including language use, neutral networks, complexity theory, self- improvement, abstractions and creativity. Although the Dartmouth conference was not very successful, it did bring the computer scientists together and established the groundwork for the future of AI research.

In 1957, the first AI program, General Problem Solver (GPS), was invented. It was able to solve a lot of common sense problems. After two years of GPS, IBM established an expert team to research AI. In 1963, MIT acquired 2.2 million dollars from the U.S. government to be used in researching Machine-Aided cognition (AI). This project attracted many computer scientists around the world to research AI technology. During the 1970's, David Mar proposed a theory about machine vision. It was a machine that could recognize objects with different shapes, color edges and textures.

During the 1980's, more advanced expert systems were developed. Digital Electronics and Xcon, which were expert systems design companies, developed programs running on the VAX computer. The demand for AI decreased in 1986-87 because most AI robot projects performed unsuccessfully. During the 1990's, AI systems were used in the military in missiles and heads-up-displays. Moreover, fuzzy logic and voice recognition systems became more available, so Artificial Intelligence technology will continue to affect our lives.

What is Artificial Intelligence?

Artificial intelligence is a pioneer of “computer science and engineering of making intelligence machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable” (John McCarthy, 2000). Moreover, AI machines can interact with humans’ thought, natural language, response and expression of emotions.

Some people think that one person can make a better response to a stimulus than another; this is defined as that person being smarter and more intelligent. Is this true? It depends on how we define intelligence. Some people define intelligence as ”comparative level of performance of a system in reaching its own objective” (Intelligent Systems and Their Societies, 1998).

Definition of Intelligence

Some people think that one person can make a better response to a stimulus than another; this is defined as that person being smarter and more intelligent. Is this true? It depends on how we define intelligence. Some people define intelligence as ”comparative level of performance of a system in reaching its own objective” (Intelligent Systems and Their Societies, 1998).

Nowadays, people often use Intelligence Quotient Scores (IQ score) to measure human mental abilities and their intelligence. Candidates need to finish a few groups of specific questions in a certain time. The test measures the candidate’s organizing and logical thinking abilities. The higher their score the more intelligent they are.

Generally, intelligence is built up from the following items and elements: (Fritz, May17, 1998; Kasper, 1998).

1.      It is established from the life experiences, for instance people can get some experiences after they make mistakes and then they learn to avoid making the same mistakes.

2.      People can accumulate different knowledge and skills and apply those to their life.

3.      People can use different ways to get insight into the problem and will select the best method to solve it.

4.      People have the ability to analyze the problem and then make a complicated situation simpler.

In order to create an intelligent machine, scientists have to know how the human brain works and what its structure is. Neural Networks are the main system in the human brain.

Neural Networks

Neural Networks are the communication structure between the human brain and the whole body. The human brain is made up of a billion brain cells and complicated neurons. The function of the neuron is to transmit the electrical signals throughout the brain and body. When a body gets a stimulus from the outside world, the brain will respond to it though the neurons immediately. For instance, if your hand gets burned by a hot cup of coffee, you must act in response to put down the cup.

Warren McCulloch and Walter Pitts proposed the theory of neutral network working on the human brain. They found that neurons work in the pattern of the binary number system (represented as 1’s and 0’s or true and false) (Methods used to create intelligence, n.d.) This is a basic electronic computer data processing pattern. Therefore, they linked this theory to the computer simulated neural network. McCulloch and Pitts designed an electronic replica, which used the neural network to prove it could perform logical processes.

The development of this project has been inhibited because of financial and technical problems, but it really has had good results, such as machines that could mimic human thinking and recognize letters.

Expert systems

An Expert systems "knowledge engineer" is a complex computer program in a specific area or field. It also embodies the knowledge to fulfill some task. The first expert system was MYCIN in 1974, which diagnosed bacterial infections of blood and suggested treatment according to a huge database. It was useful for medical students and practicing doctors in studying and for reference. Actually, the function and performance of an expert system depends on how the patient interacts with the machine. Moreover, Expert systems can also perform in other different domains such as chess, speech recognition and computer vision (McCarthy, 2000).

The COG Project:

COG is an upper-torso humanoid robot, which was created by the five researchers named Rodney A. Brooks, Cynthia Breazeal, Matthew Marjanovic, Brian Scassellati, and Matthew M. Williamson in the MIT Artificial Intelligence Lab.

COG has a head, two arms, and two eyes which are composed of four cameras and a lot of sensors and motors. Moreover, it has over twenty actuated joints, so that it has twenty –one degrees of freedom motion. COG can simulate and learn the human motion from its interaction with people. Furthermore, it can pull out a physics textbook and work out the interacting kinematics by hand. COG can learn and analyze some commands that it receives from the sensors and then it controls the motors to make responses and movements. The goal of the COG project is to design a robot which is able to interact with the world, including both objects and humans. Moreover it is used to study the theories of cognitive science and artificial intelligence. In the future, researchers will concentrate on developmental organization, social interaction, embodiment and physical coupling and multi-model integration for improving COG’s function and performance (MIT AI Lab, n.d.) In addition, another MIT research team worked on the Kismet project in order to develop a sociable humanoid robot.


Kismet is a headed robot created by the MIT Artificial Intelligence Lab; it is designed for social interaction with people. This project is mainly used to develop the emotion and human behavior on the computer. The Kismet system consists of six sub-systems: the low level feature extraction system, the high level perception system, the attention system, the motivation system, the behavior system, and the motor system.

The low level feature extraction system is used to collect the signal from outside interaction, such as sound, color and image. Afterward it sends those signals to the attention system to direct the robot’s attention and gaze toward low-level perceptual stimuli. The perceptual system is used to process the emotion and behavior response from the attention system. Afterward, the motivation system makes the robot have an intention from its "needs" and "emotions," it contains anger, disgust, fear, joy, sorrow, and surprise. The Behavior system is used to handle task-based behavior in logical structure. It makes the robot have reasonable emotion in processing the task. The main responsibility of the behavior system is to fulfill the arbitration. The motor system is capable of making the robot’s movement smoother and more expressive. The system controls the motion of the robot, for example body posture, gaze direction and vocalization (MIT AI Lab, n.d.).

My Response:

In this research project, I learned a lot of knowledge and information about Artificial Intelligence. There are some things that I never knew before such as, Neural Network, COG and Kismet. Sometime, I envision the future world from the movies and TV. Those people can communicate with the machines and make friends with them. The technology has grown faster and faster and we have achieved many goals in order to get our desires and objectives. We always like to make something possible in an impossible situation.

Although AI technology is very useful, it can also produce a lot of problems. People will depend on the computer and machines. Most jobs and operations will be more automatic, so people will face loss of their job or need to compete with the machines. Solving those problems depends on how we apply this technology to our life. For example, we can apply AI technology in medical treatment. Experts use robot arms or legs help patients working and walking.


In the future, Artificial Intelligence technology will still play a main role and be the pioneer in computer science. AI researchers will keep going to build and construct a humanoid robot and to develop expert systems that are able to understand human cognition. Actually, AI research projects will not only focus on computer science, but will also involve other sciences such as psychology, ethnology, systems theory, philosophy, and linguistics. Therefore, AI technology will be used in different places and in different ways.

Artificial Intelligence will increase our standard of living enormously and also help people to solve more complicated problems. In addition, AI technology will need a long time to develop and be researched, but then AI machines will be able to respond to people quickly and in real time.


AI on the Web. (9/28/2000). [Online].

Fritz, Walter. (5/17/1998). Intelligent Systems and Their Societies. [Online].

Kasper, Loretta F. (1998). Interdisciplinary English. New York: McGraw-Hill.

McCarthy John. (4/4/2000). What is Artificial Intelligence? [Online].

Methods used to create intelligence. (n.d.) [Online].

MIT AI Lab. (n..d.). COG Project. [Online].

MIT AI Lab. (n.d.). Kismet Project. [Online].

Return to the Top of this page

Return to Fall, 2000 Start Page

Return to ESL 91 Home Page

This page was last updated on December 2, 2000