News classification
Contact us
- Add: No. 9, North Fourth Ring Road, Haidian District, Beijing. It mainly includes face recognition, living detection, ID card recognition, bank card recognition, business card recognition, license plate recognition, OCR recognition, and intelligent recognition technology.
- Tel: 13146317170 廖经理
- Fax:
- Email: 398017534@qq.com
Artificial intelligence software will be more intelligent
Artificial intelligence software will be more intelligent
In recent years, the field of artificial intelligence has made significant progress, which has led some to fear that the software will replace people in some ways and take over work such as driving trucks. The latest reports from industry veteran scientists say they have developed artificial intelligence software that can take on the most complex part of their job designing machine learning software. Many companies, including Google, believe that learning to learn software may be able to take over some of the work of artificial intelligence experts. In one experiment, the scientists at the Google Brain artificial intelligence team let the software design a machine learning system that allowed it to be tested against language processing software, which showed that its performance was greater than previously reported Software designed by humans.
Coincidentally, in recent months, several research teams have reported on their progress in learning software for learning software from OpenAI, a non-profit research organization funded by Elon Musk, the MIT (MIT) ), The University of California at Berkeley and Google another artificial intelligence team DeepMind.
For example, the DeepMind team allowed the software manufacturing learning system to collect several different but related problems, such as walking the maze. The proposed design of the software shows that it can generalize the problem and can solve the new task with less training. The Google Brain team's research describes their software to design an image recognition system that is comparable to the best system of human design.
The team led by Otkrist Gupta, a researcher at MIT Media Lab, found that their learning software designed the performance of an in-depth learning system in object-recognition-standard testing, System compared favorably.
Self-starting AI will speed up machine learning
If this self-starting artificial intelligence technology becomes a reality, they could speed up the development of machine learning software, the researchers said. At present, the company must pay high wages to them because of machine learning experts in short supply.
"At the moment, machine learning specialists are able to solve problems because they have a professional," says Jeff Dean, head of the Google Brain Research team at the AI Frontiers Conference in Santa Clara, Calif. Knowledge, data and computing power, then do we not need so many machine learning specialists? "
Dean called the technology "automatic machine learning" and called it the most promising research direction for his team. He cautioned that, with the help of new technology, some of the work of artificial intelligence experts may be replaced by software.
However, Yoshua Bengio of the University of Montreal, Canada, stressed that, for the time being, this technology requires frightening computing power, replacing or partially replacing machine learning specialists is not realistic. Researchers say they need to use 800 high-performance graphics processors to run the software, according to researchers at Google's Brain team.
MIT Media Lab researcher Ott Cristo Gupta believes that the situation will change. He and his colleagues plan to open the source code of their own experimental software.
Google's DeepMind team calls this technology "learning tolearn," and a group of experiments show that it also helps alleviate the need for machine-learning software to call large amounts of data in order to perform well in a given task.
Thanks to advances in technology, scientists have long wished to create software to learn, but previous experiments did not produce comparable results to human-designed software. "This is exciting research," says Yoshua Bengio, who explored the idea in the 1990s.
Benxiu explained that the more powerful computing power and the recent deep concern of the emergence of learning technology, so that the latest technology possible.
Coincidentally, in recent months, several research teams have reported on their progress in learning software for learning software from OpenAI, a non-profit research organization funded by Elon Musk, the MIT (MIT) ), The University of California at Berkeley and Google another artificial intelligence team DeepMind.
For example, the DeepMind team allowed the software manufacturing learning system to collect several different but related problems, such as walking the maze. The proposed design of the software shows that it can generalize the problem and can solve the new task with less training. The Google Brain team's research describes their software to design an image recognition system that is comparable to the best system of human design.
The team led by Otkrist Gupta, a researcher at MIT Media Lab, found that their learning software designed the performance of an in-depth learning system in object-recognition-standard testing, System compared favorably.
Self-starting AI will speed up machine learning
If this self-starting artificial intelligence technology becomes a reality, they could speed up the development of machine learning software, the researchers said. At present, the company must pay high wages to them because of machine learning experts in short supply.
"At the moment, machine learning specialists are able to solve problems because they have a professional," says Jeff Dean, head of the Google Brain Research team at the AI Frontiers Conference in Santa Clara, Calif. Knowledge, data and computing power, then do we not need so many machine learning specialists? "
Dean called the technology "automatic machine learning" and called it the most promising research direction for his team. He cautioned that, with the help of new technology, some of the work of artificial intelligence experts may be replaced by software.
However, Yoshua Bengio of the University of Montreal, Canada, stressed that, for the time being, this technology requires frightening computing power, replacing or partially replacing machine learning specialists is not realistic. Researchers say they need to use 800 high-performance graphics processors to run the software, according to researchers at Google's Brain team.
MIT Media Lab researcher Ott Cristo Gupta believes that the situation will change. He and his colleagues plan to open the source code of their own experimental software.
Google's DeepMind team calls this technology "learning tolearn," and a group of experiments show that it also helps alleviate the need for machine-learning software to call large amounts of data in order to perform well in a given task.
Thanks to advances in technology, scientists have long wished to create software to learn, but previous experiments did not produce comparable results to human-designed software. "This is exciting research," says Yoshua Bengio, who explored the idea in the 1990s.
Benxiu explained that the more powerful computing power and the recent deep concern of the emergence of learning technology, so that the latest technology possible.