Machines That Can Learn Like Humans: How Artificial Intelligence is Gaining Ground
Scientists may have created a machine that can learn like a human. They've developed an algorithm that captures our learning abilities, enabling computers to recognize and draw simple visual concepts that are mostly indistinguishable from those created by humans.
"Our results show that by reverse engineering how people think about a problem, we can develop better algorithms," said Brenden Lake, one of the researchers, in a news release. "Moreover, this work points to promising methods to narrow the gap for other machine learning tasks."
When humans are exposed to a new concept, they often only need a few examples to understand it and recognize new instances of it. For example, they can understand quickly how to use a new piece of kitchen equipment or learn a new dance move. While machines can now replicate some pattern-recognition tasks previously done only by humans, machines typically need to be given hundreds or thousands of examples to perform with similar accuracy.
In order to create a machine that can learn like a human, the researchers developed a "Bayesian Program Learning (BPL) framework, where concepts are represented as simple computer programs. For instance, the letter "A" is represented by computer code, resembling the work of a computer programmer, that generates examples of that letter when the code is run. However, no programmer is required during the learning process. The algorithm simply programs itself by constructing code to produce the letter it sees. Also, every time they run these programs produce different outputs at each execution, which allows them to capture the way instances of a concept vary. In this case, it would find the differences between how two people draw the letter "A."
The BPL approach learns "generative models" of processes in the world. This makes learning a matter of model building or explaining the data provided to the algorithm. In the case of writing and recognizing letters, BPL is designed to capture both the causal and compositional process of real-world processes, allowing the algorithm to use the data more efficiently.
"We are still far from building machines as smart as a human child, but this is the first time we have had a machine able to learn and use a large class of real-world concepts-even simple visual concepts such as handwritten characters-in ways that are hard to tell apart from humans," said Joshua Tenenbaum, one of the researchers, in a news release.
The findings are published in the journal Science.