Artificial intelligence learns to create another artificial intelligence, replacing people in the future

Artificial intelligence has learned how to create another artificial intelligence and people are about to become superfluous. Automated machines can replace people in the process of creating their own " kind ".

The process of developing AI artificial intelligence causes many people to worry about the fate of themselves as well as humans that future human work will be " replaced " by automated machines. Now the main researchers realize that they can create software that is capable of learning the most difficult parts of human work - that is the task of designing machine-learning software, automated analysis will automation of building analytical models.

Artificial intelligence learns to create another artificial intelligence, replacing people in the future Picture 1

Along with a breakthrough in technology, which means bad news for those concerned, leading scientists are investing in ways to allow software to create a machine software. -Learning other. They are on their way to finding a revolutionary automation software for AI.

In an experiment, scientists at the artificial intelligence research group Google Brain used a machine-learning system design software for the purpose of testing the benchmark capability of another language processing system. The results show that the new software outperforms the old software designed by humans.

In recent months, several other research groups have also provided information about the process of " creating software to create other software ". The above groups include members from the OpenAI nonprofit research organization ( co-founded by Elon Musk ), Massachusetts Institute of Technology (MIT ), University of California, Berkeley and DeepMind's research team. Google.

Artificial intelligence learns to create another artificial intelligence, replacing people in the future Picture 2

If this method of artificial intelligence creation is widely applicable, it can speed up the process of making machine-learning software economically fast. Currently, the cost of hiring machine-learning professionals is not cheap, if the automated machine can do the human part, even in the making of its own " fellow ", it can be human. will become an excess factor in the AI ​​manufacturing cycle. Because now companies are paying insurance for machine-learning professionals, the position is lacking staff.

Jeff Dean, head of research at Google Brain, said in a statement last weekend: " Workers in some production stages may be more effectively replaced by a software ." Jeff Dean also said that " automatic learning learning " technology is one of the most promising research projects his team is currently investing.

" Currently, how to solve problems including expertise, information and calculations. Can we completely eliminate expert opinions about machine-learning? " Dean said at the AI ​​conference. Frontiers in Santa Clara, California.

Artificial intelligence learns to create another artificial intelligence, replacing people in the future Picture 3

A series of experiments from Google's DeepMind team suggests that the " method to learn to learn" that researchers are applying will reduce the huge amount of data that a software- learning needs to be able to operate in the most efficient way " .

The researchers challenged their own software, demanding to create a learning system to collect all different issues, but related to a major goal, thereby creating a system design. new in that design. They see the ability to reproduce, select new tasks without having to go through the usual preparation steps today.

The idea of ​​creating "learning to learn" software is not so new, but previous tests often do not yield the desired result: " They are not well-suited to human designs ". However, this is still considered a potential aspect of the artificial intelligence development industry, Professor Yoshua Bengio from the University of Montreal said " It is very interesting " to study this idea in the 1990.

Professor Yoshua Bengio said: " The existing computer systems are becoming increasingly powerful and with technology called deep learning - what is interesting recently about AI artificial intelligence is what is doing the research. rescue system approach "learn to" have a strong potential to rise up ". In addition, the professor added: " Such a system will need a tremendous computing power to replace human experts in this area ."

Artificial intelligence learns to create another artificial intelligence, replacing people in the future Picture 4

Researchers at Google Brain also describe a powerful system that uses 800 graphics processors to power the software, thereby creating an equal image recognition system ( and even pass ) design created by people.

Dr. Otkrist Gupta, an MIT Media Lab researcher, believes that AI manufacturing will soon change. He and his colleagues at MIT planned on an open source software, where the learning software will design itself so that a deep-learning system can identify strong and accurate images on par with a system created by humans.

Dr. Gupta was inspired to carry out the project by spending a lot of time designing and testing machine-learning models. He said that companies and research also have the motivation to design computers that automate machine learning.

" Reducing the burden on scientists will be an effective solution. It can make us more productive, create more efficient system templates and give us free time to explore those. The idea is higher, "said Dr Gupta.

Refer to some more articles:

1. What happens if aliens are artificial intelligence?
2. Top 10 fastest growing industries in 2016
3. 35 inventions that make your life easier

Having fun!