Significant advancements have been made in the field of machine learning but we’re still not at a point yet where we can say that a machine is finally able to match a human’s ability to see one example and intuitively figure out what a symbol or object might mean. Computers first have been fed countless examples before they’re able to make a similar observation, but a new algorithm developed by researchers from New York University may enable machines to take mental leaps like humans when they’re learning new things.
That’s not all, the algorithm also enables the machine to recreate simple symbols and drawings that’s not unlike the way a human would draw those symbols and drawings. This research which describes the creation of the Bayesian Program Learning has been published in Science, basically it turns a concept into a computer program, enabling computers to learn new things with a single example.
The model can also use knowledge from previous concepts to figure out how something works, like when a computer already knows a Latin alphabet, it can easily learn a similar Greek alphabet.
It’s interesting to note that when the machine was told to create new examples based on the concept it was taught, it created examples that were similar to the ones that humans created, other humans couldn’t even distinguish whether the examples had been created by a human or a machine.
Computers will thus be able to learn new things much quickly, effectively adapt to new situations and scenarios without first having to be taught thousands of examples. Realistically though applications for this research may take years but it’s good to know that work on advancing machine learning is going in the right direction.