Google's CEO, Sundar Pichai, spoke of his excitement about having AI build AI, which he believes will bring on the next wave in the fourth industrial revolution.
Experts in AI are few and far between, and most of the world's foremost businesses are crying out for them. We're not only talking about the tech companies like Google, Microsoft, Apple and Amazon, but the ones you wouldn't associate with technology as well. Legal firms, insurance companies, finance businesses... They all want to delve into the world of AI, but that means having to find people who understand it and build the platforms they need.
However, this may all be about to change, especially if Google's CEO, Sundar Pichai, is to be believed. He said that one way to breach the AI skills gap is to have machine-learning software take on some of the work in creating other machine-learning software. In short, AI building AI. Sounds scary, but it could actually solve a whole slew of issues in the market.
At Google's annual developer conference, I/O, Pichai announced a new project referred to as AutoML, which is being developed by Google Brain's AI research and development group. They have worked out a way to get their algorithms to automate some of the toughest and most time-consuming parts of the task of building machine learning software. In a few instances, Google's new automated system developed AI equal to or even better than what the human experts were able to create, which is outstanding, and game-changing.
"This is a very exciting development,” Pichai told MIT Technology Review. “It could accelerate the whole field and help us tackle some of the most challenging problems we face today.”
Pichai has high hopes of the AutoML project. He envisages a future in which developers' time is spent on doing the mundane and time-consuming tasks, all of which will be left to AI. He also hopes that it'll expand the number of developers who are capable of making use of machine learning without needed the high-end expertise.
This approach would also be in lone with Google's attempts to make their cloud computing service the best on the market with regards to machine learning. Google is also trying to pave its path into the corporate cloud computing world, which is currently dominated my Amazon and Microsoft, in that order.
AutoML will also help make deploying an AI-based technique called deep learning, which is used to power AlphaGo. Deep learning is also used to help power speech and image recognition, translation, and in some cases, robotics.
Researchers across the globe are testing the boundaries of deep learning. Researchers at Deep Mind, at Universities, and Elon Musk's OpenAI are exploring the possibilities of this nascent and relatively unknown technology.
Yet will those developing AutoML be about to put themselves out of a job? When two of the researchers behind the project, Barret Zoph and Quoc Le, were asked this question, they laughed and claimed that as it stands, the system is far too expensive to be used by businesses. Then you have to factor in the amount of energy that these experiments have used. Google's energy bills must be through the roof.
Image Credit: Nguyen Hung Vu