Image via Shutterstock
Can an artificial brain truly replicate the workings of a human mind? This week, the final Reddit posts from renowned physicist Stephen Hawking made their rounds on the internet. Hawking’s stance towards AI was that it’s possible for robots to outsmart humans, just like how “Einstein was smarter than his parents.”
Now, a team of scientists at the nonprofit Allen Institute for AI (Ai2) has sought to figure out if AI is capable of “common sense,” which helps humans to fill in gaps when they’re faced with a question.
The researchers have created a test called the ‘Arc Reasoning Challenge’ (ARC) to determine if AI can understand the world like humans do.
Here, computers are asked a series of general knowledge multiple choice questions like, “Which item below is not made from a material grown in nature?” Answers include “a cotton shirt… a wooden chair… a plastic spoon… [and] a grass basket.”
A layman would immediately point to the plastic spoon, but a machine might struggle with an answer. AI is often trained on specific data to, for instance, stand in as IT technical support—but pose a question that requires broader knowledge and it would likely fail.
“Machines do not have this common sense, and thus only see what is explicitly written, and miss the many implications and assumptions that underlie a piece of text,” concluded Peter Clark, who leads the ARC project.
Should a computer pass the ARC test, it would be able to comprehend general knowledge, a skill that no AI has obtained yet.