Researchers Quantify 'Common Sense' to Eventually Teach AI
-
Researchers from UPenn have developed a framework to quantify "common sense" in order to eventually incorporate it into AI systems. They define it based on the level of consensus around a given claim.
-
The claims that were most "commonsensical" were plainly worded facts about the physical world. People with greater awareness of others' emotions exhibited more common sense.
-
Quantifying common sense is important for developing AI tools that actually have common sense rather than just treating it as "obviously true" things.
-
Even though individuals may share many beliefs, as a group gets larger the portion of shared beliefs drops quickly. This means people likely understand things differently than we assume.
-
The research has implications for business leadership - leaders can't assume people understand things the same way they do. Everyone has some "common sense" but it's still unique.