Research Group Projects and Descriptions

The Society of Mind research group focuses on imparting to machines the human capacity for commonsense reasoning. We account for many aspects of commonsense, imagination, and reasoning by analogy as resulting from negotiations among different cognitive processes that use different ways of representing knowledge.

AnalogySpace enables common-sense reasoning through principal component analysis. It projects the information in ConceptNet into a reduced-dimensional space that describes common-sense concepts and their properties in terms of automatically discovered correlations called "eigenconcepts." AnalogySpace can be used to infer new information, reason about ad hoc categories, detect topics in text, and compare concepts on scales that can be generated on the fly.

We are developing next-generation architectures for artificial intelligence based on Professor Minsky's "Society of Mind" theory of human thinking. The main idea is that the key to human flexibility and resourcefulness is mental diversity: we have many ways to solve every kind of problem; when we get stuck trying one method of solution, we can switch to another. We are exploring how this idea can be applied at different places and levels in a cognitive architecture, in order to build systems capable of robust common-sense reasoning.

Alumni Contributor(s): Push Singh

Imparting common-sense knowledge to computers enables a new class of intelligent applications better equipped to make sense of the everyday world and assist people with everyday tasks. While previous attempts have been made to acquire and structure common-sense knowledge, they have either been inadequate in capturing the breadth of knowledge needed for the enterprise, or their complicated representation schemes have made them difficult to incorporate into applications. Our approach to this problem is ConceptNet, a freely available common-sense knowledge base that possesses a great breadth of knowledge that can be easily incorporated into applications. Built from the Open Mind Common Sense corpus, which acquires common-sense knowledge from a Web-based community of instructors, ConceptNet is a semantic network of 1.6 million items of common-sense knowledge, and a set of tools for making inferences using this knowledge.

Alumni Contributor(s): Xinyu H. Liu and Push Singh

Funk2 is a novel process description language that keeps track of everything that it does. Remembering these causal execution traces allows parallel threads to reflect, recognize, and react to the history and status of other threads. Novel forms of complex, adaptive, nonlinear control algorithms can be written in the Funk2 programming language. Currently, Funk2 is implemented to take advantage of distributed grid processors consisting of a heterogeneous network of computers, so that hundreds of thousands of parallel threads can be run concurrently, each using many gigabytes of memory. Funk2 is inspired by Marvin Minsky's Critic-Selector theory of human cognitive reflection, and is the foundation for the Neural Models of Mind project.

LifeNet is a probabilistic spatial and temporal model allowing common-sense physical simulation. Mixtures of gaussians are used to map from the physical world's real numbers to human language's symbolic conceptual world. Humans are very good at moving around and physically manipulating the world in which they live, quickly considering many collections of physical events and objects in order to choose a set of actions. The current state of the art in computer physics simulations does not take advantage of common-sense knowledge such as "things usually fall if not supported." Building a common-sense physics simulation allows us to take a paragraph of text and quickly reconstruct the physical situation in more or less detail as required for inference, planning, and further reflective algorithms. Immediately possible applications include interpreting sensor data, comparing physical descriptions of events in text, and efficiently planning in multiscale, robotic physical manipulation environments.

Alumni Contributor(s): Push Singh

This project addresses human cognitive models of reflective problem solving in terms of psychology, neuroscience, and artificial intelligence. A programming language describing reflective human thought processes is being developed for the purpose of understanding the biological process of thought. This description language allows distributed reflective monitoring and control of parallel threads. In addition to being a novel method for the robust control of distributed computer programs, this technology is directed toward consumer HCI and medical cures for neuropsychological problems, and has applications for neural-interface computer gaming peripherals, aging population cognitive evaluation, and training.

The biggest problem facing artificial intelligence today is how to teach computers enough about the everyday world so that they can reason about it like we do—so that they can develop "common sense." We think this problem may be solved by harnessing the knowledge of people on the Internet, and we have built a Web site to make it easy and fun for people to work together to give computers the millions of little pieces of ordinary knowledge that constitute "common sense." Teaching computers how to describe and reason about the world will give us exactly the technology we need to take the Internet to the next level, from a giant repository of Web pages to a new state where it can think about all the knowledge it contains; in essence, to make it a living entity.

Alumni Contributor(s): Push Singh

The Open Mind Common Sense project has collected hundreds of thousands of statements of common-sense knowledge from volunteers on the Internet, using a variety of online activities in several different languages. Open Mind Commons aims to use analogical reasoning to make connections between similar ideas while highlighting the relevant differences as well. These analogies can give a computer a better understanding of the relationships between objects, situations, and cultures. It is often difficult to search through and coordinate lexical information across data sources, each of which has its own separate interface and viewing software. We have approached this problem by creating a unified, flexible interface for various natural-language processing resources.

Alumni Contributor(s): Hyemin Chung

Roboverse is a physical artificial intelligence simulation. The environment supports rigid body physical simulation of wheeled and legged robots with default algorithms for handling the basic reactive control layers to move from one location to another by straight line, to observe objects in the local neighborhood of the robot's 2D position, to pick up, translate and rotate objects by using a servo interface. Higher level cognitive functions, such as speech and social problem solving were developed as part of Push Singh's cognitive architecture design, the Emotion Machine v1.0 (EM-1).

Alumni Contributor(s): Push Singh

Professor Minsky continues to develop the theory of human thinking and learning called the "Society of Mind," which tries to explain how various phenomena of mind emerge from the interactions among many different kinds of highly evolved brain mechanisms. In this way we can account for many aspects of common sense, imagination, and reasoning by analogy, as resulting from negotiations among systems that use different ways of representing knowledge. Similarly, it appears that we can explain many of the regularities found in natural languages as consequences of how those representations work, rather than as constraints that are externally imposed on interpersonal communications. This approach also suggests that some of what we call "emotions" are mechanisms required for managing conflicts among competing goals. We may need to construct similar systems when we begin to build smarter and more versatile machines.