Is It Possible for Computers to Learn Like Humans?
Recent technological advancements have ushered in a new age of digital natives. With technology aiding almost every facet of daily life, it’s almost impossible to imagine a day without it. Technology involving artificial intelligence (AI), for instance, have capabilities, such as driving cars or picking out movies.
As AI grows to be more sophisticated, one question comes to mind: is it possible for computers equipped with AI to learn, or even reason, like humans?
Machine Learning and Deep Learning
The key to AI is machine learning. Here, the user writes a program that lets the computer learn a task instead of telling it how to do something. In particular, machine learning focuses on developing computer programs that can access data and use it to learn for themselves. These programs mainly aim to allow computers to learn automatically without human intervention or assistance.
Say, for instance, you’d like to have your computer pick out individual voices from a crowded room. Nobody can properly explain how the human brain does this, making it difficult to tell computers how to achieve this task. Experts, however, are able to impart this capability to computers through deep learning.
How does it work? First, you provide the input to the computer, along with the transcript of what people are saying. From meaningless noise, the computer will then learn what sounds go with what words. Through practice, the computer can then apply what it has learned to sounds it has never heard before.
AI and Relational Reasoning
Relational reasoning is a requirement in making everyday decisions, like where to go to dinner or what to buy. This is something that AI is yet to be capable of – or so we thought.
Researchers at Google’s DeepMind were able to come up with an algorithm that can handle such reasoning, and is already seen to be better than humans at a complex image comprehension test. The algorithm proposes an artificial neural network for relational reasoning that stitch together tiny programs that will collaboratively find patterns in data. Basically, the researchers are “forcing the network to discover the relationships that exist between the objects.”
Such an algorithm, however, still needs to learn how to answer more challenging questions to approach human-like flexibility. This means comparing not just pairs of things, but triplets, pairs of pairs, or something similar.
While AI is still miles from having the same learning and reasoning capabilities that humans have, recent developments are enabling the emergence of tools that are as smart and capable as everyone.