
The exponential progress in the field of robotics has already been feared to take so many production jobs away from humans, and the latest edition to those victims might be the programmers. Researchers at the Google Brain artificial intelligence have designed a machine learning system that can develop machine learning software.
Interestingly, when compared, it exceeded the results from the ones designed by humans.
According to Jeff Dean, who leads the Google Brain research group, such exertion could supplant some of the work from the workers and enhance the pace of the implementation of the AI software in different fields of economy.
“Currently the way you solve problems is you have expertise and data and computation,” said Dean, at the AI Frontiers conference in Santa Clara, California. “Can we eliminate the need for a lot of machine-learning expertise?”
The researchers challenged their software to create a machine learning system, which they are terming “learning to learn.” Such systems are believed to reduce the need for huge amounts of data used by machine learning software in order to perform the task well.
The researchers used 800 high-powered GPUs to power the software that showed an ability to generalize and pick new tasks. They created machine learning systems for different related problems using their software.