Share Site Map Contact us Home Page
Home > Background > AI at the turn of the millennium > Academic AI

History of AI

AI at the turn of the millennium
Academic AI
Corporate AI
Commercial AI

Speaking Machines

From Myth to Reality part 1

From Myth to Reality part 2

From Myth to Reality part 3

From Myth to Reality part 4

From Myth to Reality Full
 Academic AI
The MIT AI Lab
Harvard University, Computer Science Department
Carnegie Mellon University's Human-Computer Interaction Institute
Princeton University - School of Engineering and Applied Science
Computer Science Department - Stanford University
The SRI International Artificial Intelligence Center (AIC)
The Artificial Intelligence Group at the Beckman Institute - University of Illinois
The Information Science Institute at the University Of Southern California
AI Laboratory - University of Michigan
The Interdisciplinary Center for Neural Computation at the Hebrew University
MIT Media Lab - Software Agents Group
The MIT AI Lab
There are several major projects at the MIT AI Lab, each focused on different approaches to building artificial intelligence. The COG project, for instance, seek to build a humanoid robot; the InfoLab Group is looking for ways to manage databases information using natural language.

Taken as a whole, the goal of the MIT AI Lab is to create a computational theory of intelligence that covers everything from how behavior works to the operations of neurons. The lab does research into the roles of vision, language, and motor computation in general intelligence. The lab also seeks to develop high-impact practical applications in areas such as:

  • information transportation,
  • enhanced reality,
  • the human-computer interface,
  • modeling and simulation,
  • image understanding,
  • decision analysis and impact prediction,
  • model-based engineering design,
  • and medicine.

The COG Project: Headed by Prof. Rodney Brooks, the Cog Project is building a humanoid robot. The robot is composed of a set of sensors and actuators which, taken together, are designed to simulate the sensory and motor dynamics of a human body.

The COG project uses behavior-based reasoning to control these disparate elements. This approach views intelligence not as a single thread of thought, but as the result of the interactions of many simple components, all working at the same time.

In contrast to neural networks or parallel distributed processing, the units of intelligence are not just homogenous equations, but are manifestations of recognizable behaviors, linked in particular ways. Another theory employed at MIT is reactive reasoning, which relates to the performance of actions without "planning."

In recent years, the term "reactive" has come to be employed for any action triggered by reacting to the environment rather than by deliberation, or cognitive assessment. Reactive, behavior-based intelligence is composed of many interacting reactive "sub-behaviors." The application of these theories has resulted in the creation of reactive robots that can reliably interact with the real world at realistic, animal-like speeds. However, over a decade has passed since the approach was first introduced, it is still not clear whether it can lead to robots capable of more sophisticated behavior.

The InfoLab Group: The InfoLab group develops interactive software systems that help people access information and solve problems using natural language. The group conducts research into natural language processing and multimedia information access.

A key component of this effort is the START Natural Language System developed by Boris Katz. START provides access to text, diagrams, images, and other types of information in response to questions posed to it in English. It uses a technique called "natural language annotation," in which retrievable information segments are associated with normal sentences and phrases which describe their contents. These information segments are then retrieved when their annotations semantically match particular user-submitted questions. Another component of the group's effort involves the development of IMPACT, an interactive planning and monitoring system grounded in simple English accounts of circumstances and events.

Hal (The Intelligent Room Project): Hal is a highly interactive environment that uses "embedded" computation to observe - and participate in - the everyday events happening around it. An offshoot of the MIT AI Lab's Intelligent Room, Hal has cameras for eyes, microphones for ears, and uses a variety of vision, speech and gesture recognition syste
Harvard University, Computer Science Department
The research group, headed by Professor Barbara Grosz, is developing systems ("agents") able to collaborate with each other and their users. Professor Grosz is also attempting to identify the basic structures and processes of language use, focusing on the mechanisms involved in dialogue and spontaneous speech.
Carnegie Mellon University's Human-Computer Interaction Institute
Carnegie Mellon University's (CMU) Human-Computer Interaction Institute (HCII) is an interdisciplinary community of students, faculty, government, and industrial companies, working together on a wide spectrum of human-computer applications.

Current projects at CMU include:
  • user interfaces for hands-free operation,
  • intelligent tutoring systems,
  • digital video computer systems,
  • computer languages,
  • automatic speech processing and assessment of speech recognition systems.

The Command Post of The Future Project: This Army-funded project seeks to create the ultimate command post - an environment that is totally reactive to the users. Individuals will be able to make use of large amounts of linked information via speech recognition, gesture recognition, and PDAs; information will be updated constantly.

The Demonstrational Interfaces Group: The Demonstrational Interfaces group wants to improve human-computer interaction by making the computer a better learner. With a demonstrational interface, you, as a user, give the computer an example of how you'd like it to operate - and it generalizes your example into a process that you can repeat as often as possible. Demonstrational interfaces are a way to avoid repeating the same procedures - like formatting certain kinds of text, or drawing plans or graphs - over and over again.

Engineering Models of Human Performance: This groups looks for ways to predict how people use computers in different situations, which allows software engineers to create programs which make sense to the user and save time. According to the group's web site, a tool they developed, called CPM-GOMS, predicted how telephone operators could most efficiently use customer service software, and saved the NYNEX telephone company $2 million a year.

The Informedia Project: The Informedia project focuses on new approaches for automated video and audio indexing, navigation, search and retrieval - creating easy-to-search "knowledge bases" from moving images and recorded sound. They're developing both educational and entertainment systems using this indexing technology.
Princeton University - School of Engineering and Applied Science
The research programs at the Princeton School of Engineering cover a broad range of computer science and artificial intelligence related issues. The Computer Audio and Human Interaction group seeks to understand how computers can be best used to understand and generate speech.

Computer Science Department - Stanford University
As at MIT, artificial intelligence is a major field of study at Stanford. The Computer Science department hosts a number of related research projects, covering many different topics, including:

  • basic research in artificial intelligence and formal reasoning,
  • expert systems,
  • large knowledge bases,
  • agent-based architectures,
  • image understanding,
  • robotics,
  • machine learning,
  • mathematical theory of computation,
  • program synthesis and verification,
  • natural language understanding,
  • parallel architectures,
  • design/manufacturing,
  • and portable LISP systems.
The SRI International Artificial Intelligence Center (AIC)
The Artificial Intelligence Center (AIC) at SRI (Stanford Research Institute) is one of the world's major centers of research in artificial intelligence. Broadly, SRI seeks to understand the "computational principles underlying intelligence in man and machines," giving machines the ability to solve problems, communicate effectively with people, and interact with the natural world. Research Programs in artificial intelligence at SRI include a Natural Language Program focused on multimedia interfaces, spoken language systems, and written language systems.
The Artificial Intelligence Group at the Beckman Institute - University of Illinois
The AI Group at the Beckman Institute has two main research tracks: biological intelligence and human-computer intelligent interaction (HCII). The group looks for ways to give computers perception and awareness of their environments - and also looks at human behavior when designing intelligent systems. At the institute, research is performed on artificial intelligence, robotics, computer vision and cognitive science.
The Information Science Institute at the University Of Southern California
At the Information Science Institute (ISI), a part of the University of Southern California (USC), 325 researchers work in a broad spectrum of information processing research, attempting to develop advanced computer and communication technologies.

Most of the research group's current work has to do with Internet architecture, exploring how files are arranged over networks and implementing new ways of doing teleconferencing. As part of their overall strategy, the Natural Language group at ISI is conducting research into various areas of Computational Linguistics and Natural Language Processing, trying to find ways to allow humans to speak comfortably with computers, and also with each other, regardless of language, through automatic language translation.
AI Laboratory - University of Michigan
The Artificial Intelligence Lab at the University of Michigan is researching a variety of artificial intelligence topics, related to natural language processing, rational decision making, distributed systems of multiple agents, machine learning, cognitive modeling, collaboration technology, real-time and intelligent dynamical control, and building digital libraries.
The Interdisciplinary Center for Neural Computation at the Hebrew University
The center is studying the biology of learning to find new ways to improve machine learning and create artificially intelligent devices. The Center also researches simple local learning rules underlying complex temporal and other behavior patterns, trying to find the building blocks of behavior and cognition. At the Institute of Computer Science at the Hebrew University, Associate Professor Jeffrey Rosenschein heads a research group focused on issues of cooperation and competition among agents, and the use of economic theory, voting theory, and game theory to establish appropriate foundations for Distributed Artificial Intelligence.
MIT Media Lab - Software Agents Group
Also at MIT, the Software Agents group of the MIT Media Laboratory is looking for ways to create computer systems, called "agents," which are "semi-autonomous, proactive, and adaptive." A software agent, in theory, can keep your checkbook balanced and order pizza for you, while renegotiating your mortgage and trading your stocks. The group develops techniques and builds prototype agent systems that can be tested on the Web and in other computing environments.
  Printable version