The Story of Watson
This IBM AI system gained fame after defeating Ken Jennings and Brad Rutter on Jeopardy in 2011
What is Watson?
The comparison of true language comprehension to this kind of reasoning humans use daily to interpret context.
What is commonsense reasoning?
These deliberate small changes can fool AI systems into giving wrong answers or labels.
What are adversarial attacks (or adversarial examples)?
Stanford professor Christopher Manning argued that AI will never fully understand language without this key human ability.
What is common sense?
IBM's Watson buzzed in only when its top guess passed this internal measure of certainty
What is a confidence threshold?
These sentence pairs, invented by Terry Winograd, test whether a system truly understands pronouns and meaning in context.
What are Winograd schemas?
This Stanford dataset became a benchmark for machine reading comprehension
What is SQuAD (Stanford Question Answering Dataset)?
NLP systems, like image classifiers, can be fooled by subtle sentence edits that shift meaning, exposing their lack to this
What is true language comprehension or contextual understanding?
IBM’s marketing campaign described Watson as the dawn of this vaguely defining computing era
What is cognitive computing?
The Winograd schema test was proposed as an alternative to this famous measure of intelligence.
What is the Turing Test?
In SQuAD, answers are always found directly in the text, making it a test of this still rather than true understanding.
What is answer extraction?
Mitchell argues that machines will not fully understand language until they share this human trait
What is human experience?