Machines will be able to replace humans in the process of creating the “fellow” of them. Process development of artificial intelligence has made many people worried for their own fate as well as of men: they believe that we are facing a potential tyrant champion holding a power or gently rather, the work of people in the future will be taken off in part by automated machines. Now, with the advances of technology breakthrough means bad news for those who worry the above, when the leading researchers are studying how investment allows the software to be able to create the part other machine-learning software. They are on the road looking for a revolutionary software manufacturing AI.
In one experiment, the scientists at the research group Google Brain artificial intelligence software used to design a machine-learning system for the purpose of testing the ability of a system benchmark language processing other languages. The results showed that the new software excels old software designed by humans. A recent months, a number of research groups have also provided information on the process of “creating software created by other software” her.
The group which includes members from the research organization nonprofit OpenAI (co-founded by Elon Musk), MIT, University of California and the team DeepMind Google. If this way of building AI can be widely applied, we can accelerate the process of manufacturing machine-learning software up fast. Currently, the cost to rent a machine-learning professionals is not cheap, if machines can replace human parts including in the manufacture of the “fellow” of them, maybe people will become a redundancy factor in the fabrication cycle AI.
Jeff Dean, Google Brain study leader, said in a weekend statement that workers in some stages of production will be replaced by a software efficiency. He said that the technology “automatic machine learning” project is one of the most promising research that his team is currently investing. “Currently, the way to solve the problem including expert opinions, information and calculations,”
Dean spoke at AI Frontiers conference in Santa Clara, California. “If we can eliminate entirely the expertise of machine-learning what?”. A series of experiments from Google DeepMind group suggested that the method called “learning to learn” that the researchers are applying will reduce the huge amount of data that a machine-learning software needed to be able to operate most efficiently. They tested their software, asking it to create a learning system to collect all the issues related to a key goal, which requires it to create a new design and system design in which they have views of the regenerating, choosing new tasks without going through the usual preparation.
The idea of creating a software “learn to learn” is not so new, but previous trials often do not bring the desired result: they will not be with the packaging of human design. However, it remains to be a promising aspects of industry development in artificial intelligence, Yoshua Bengio professor from the University of Montreal have identified such a study when this idea in 1990.
He said Bengio computer systems increasingly stronger, and with a technology called deep learning, a system of “learning to learn” has the potential to rise strongly. But he added that such a system will need a computing power extremely large, to be able to replace the human experts in this field. Researchers at Google Brain also gave a description of a system so powerful. They use 800 graphics processor to deliver robust software for energy, thereby creating an image recognition system on par (and even surpass) the design created by humans.
Dr. Otkrist Gupta, a researcher at the MIT Media Lab AI believes that the manufacturing sector will soon change. He and colleagues at MIT have planned for an open source software where the software will learn to design a deep-learning system that can identify reliable and accurate images on a par with a system created by humans.
“Reducing the burden on the shoulders of scientists would be an effective solution,” he says Gupta.
“It can make us more productive, create the template system more efficient and give us free time to explore the idea of higher rank.”