Open Mind Common Sense (OMCS) is a project that aims to build and utilize a vast commonsense knowledge base from the contributions of many people across the Internet. The goal of the project is to enable AI systems to understand and interact with humans in a more natural and intuitive manner.
Imagine you’re helping to teach a very smart robot about the world. You and thousands of other people write down things you know - things that seem obvious to you, but a robot wouldn’t know. That’s what the Open Mind Common Sense project does – it collects all these ‘obvious’ facts to help robots understand us better.
Open Mind Common Sense (OMCS) is an ambitious project initiated by the MIT Media Lab in 1999. The goal was to build a large-scale common sense knowledge base by crowdsourcing information from people across the internet. OMCS sought contributions from a vast array of individuals to capture as much diversity and breadth in its knowledge base.
This common sense knowledge base is out there to serve artificial intelligence systems. Machines inherently lack the vast breadth of ‘common sense’ knowledge that humans have gathered through day-to-day living and experience. When we speak about ‘common sense’ here, we are referring to the vast array of facts, understandings, and reasoning processes which are ‘obvious’ to any person but not inherently understood by a machine.
Examples of such knowledge could be ‘people usually sleep at night’ or ‘coffee can make you feel awake’. These are assumptions or facts that a human wouldn’t typically need to articulate, whereas an AI system wouldn’t automatically know without being taught.
At a functional level, the OMCS project uses a form of representation called “ConceptNet” to store this knowledge, which is essentially a semantic network allowing various concepts to interconnect. Once acquired, this knowledge can be used in AI systems to provide more human-like reasoning capabilities and responses, improving their ability to interact naturally with humans.
The OMCS project has been used in several influential AI systems, such as the language model GPT-3, where it has been an essential source for training data contributing to the model’s impressive language understanding capabilities.
Commonsense reasoning, Knowledge base, Crowdsourcing, Semantic network, GPT-3, ConceptNet.