Artificial intelligence seems to be everywhere, but what we are really witnessing is a supervisedlearning revolution: We teach computers to see patterns, much as we teach children to read. But the futureof A.I. depends on computer systems that learn on their own, without supervision, researchers say.When a mother points to a dog and tells her baby, “Look at the doggy,” the child learns what to callthe furry four-legged friends. That is supervised learning. But when that baby stands and stumbles, againand again, until she can walk, that is something else. Computers are the same. Just as humans learn mostlythrough observation or    20    , computers will have to go beyond supervised learning to reach the holygrail of human-level intelligence.“We want to move from systems that require lots of human knowledge and human hand engineering”toward “increasingly more and more    21    systems,” said David Cox, IBM Director of the MIT-IBMWatson AI Lab. Even if a supervised learning system read all the books in the world, he noted, it wouldstill   22    human-level intelligence because so much of our knowledge is never written down.
        Supervised learning depends on annotated data: images, audio or text that is painstakingly labeledby hordes of workers. They circle people or outline bicycles on pictures of street traffic. The labeled datais fed to computer algorithms, teaching the algorithms what to look for. After ingesting millions of labeledimages, the algorithms become expert at recognizing what they have been taught to see.
        But supervised learning is constrained to relatively narrow domains defined largely by the trainingdata. “There is a limit to what you can apply supervised learning to today       23    the fact that you need alot of labeled data,” said Yann LeCun, one of the founders of the current artificial-intelligence revolutionand a recipient of the Turing Award, the equivalent of a Nobel Prize in computer science, in 2018. He isvice president and chief A.I. scientist at Facebook.
        Methods that do not rely on such precise human-provided supervision, while much less explored,have been eclipsed by the success of supervised learning and its many practical applications — from selfdriving cars to language translation.    24    supervised learning still cannot do many things that are simpleeven for toddlers.
       “It’s not going to be enough for human-level A.I.,” said Yoshua Bengio, who founded Mila, theQuebec AI Institute, and shared the Turing Award with Dr. LeCun and Geoffrey Hinton. “Humans don’tneed that much supervision.”
        Now, scientists at the forefront of artificial intelligence research have turned their attention back toless-supervised methods. “There’s self-supervised and other related ideas, like reconstructing the inputafter forcing the model to a compact representation, predicting the future of a video or masking part of theinput and trying to reconstruct it,” said Samy Bengio, Yoshua’s brother and a research scientist at Google.
Source: https://www.nytimes.com/2020/04/08/technology/ai-computers-learning-supervised-unsupervised.html?searchResultPosition=1
20.
(A) tumble and fall
(B) trial and error
(C) cloning and pirating
(D) pros and cons