Computer science is an ever changing field, and many await to see which new gadgets will be buzzed about in the tech industry.
One very hot topic is Artificial Intelligence, or AI. AI is a machine’s ability to mimic a human’s consciousness and self-awareness.
Machines can already do so much from pattern recognition and identification of facial expressions; some consider it only a matter of time before an Ai is created.
A recent paper by MIT Technology Review asked the question, What qualifies a machine to be the equivalent to human intelligence?
In attempting to answer this question, researchers looked into an AI research group from the University of Illinois Chicago.
The machine system this group uses is called ConceptNet, an open-source project created by MIT’s Common Sense Computing Initiative. ConceptNet implements a Python code module of AnalogySpace.
The MIT researchers decided to test this machine’s intelligence with a common IQ test, the Wechsler test.
The result was a score of WPPSI-III VIQ, an average score for four year olds.
The test was given verbally, with ConceptNet answering the questions it was asked.
However, the words it used were very jumbled, not proper english.
Also, it sometimes converted the meaning of one word to that of another word. Such as the words saw and see.
For example, one question the machines was asked was “What is a saw used for?”
To which the machine responded, “An eye is used to see.”
When the machine was asked questions that had to do with inherent meaning, typical with human interaction, the machine dropped in performance.
When the machine was asked “Why do people shake hands?” it responded “They are having an epileptic fit.”
All things considered, this is a large step for the AI world and computer scientists alike.
It has taken 60 years to get to this point, and progress is increasing rapidly.