Recently, the US authoritative technology media Extreme Tech published a paper on Baidu's progress in the field of artificial intelligence deep learning: Baidu's enhanced algorithm enables artificial intelligence to obtain the grounding language organization ability, enabling artificial intelligence to achieve zero sample learning. This is the most difficult part of the current development of artificial intelligence. This new achievement will advance the advancement of natural language processing, allowing machines to one day understand the language as humans do.
Difficulties in Baidu to achieve zero-sample learning in artificial intelligence language
The enhanced algorithm made by Baidu enables artificial intelligence to acquire the language organization ability of grounded air. In this regard, the US authoritative technology media Extreme Tech recently published a paper on this major advancement in the field of deep learning, Baidu artificial intelligence has successfully broken through the most difficult part of the current development of artificial intelligence: "zero sample learning."
In the article, Baidu's algorithm allows the artificial intelligence agent (Agent) in the video game to learn English, understand the written commands, and then do it through the "virtual teacher", which enables Baidu artificial intelligence to achieve zero sample learning. Learning to handle tasks without any samples is the most difficult part of the current development of artificial intelligence.
Specifically, Baidu's “zero sample learning†project allows the Agent to respond to the commands given by the teacher and then give the virtual teacher positive or negative feedback. If the agent can make the correct action, it will receive the reward, and if it fails, it will be punished. Agents can slowly understand the meaning of the language and learn to use the language correctly. Thereafter, if an unfamiliar command is encountered, the Agent can infer the correct meaning of the language to meet the expected goal, which is zero sample learning. Only the whole process takes place in a simple 2D maze, which can then be applied to the simulated 3D environment and then applied to the real scene.
Language-based use, zero-sample learning or small sample learning is not difficult for humans to understand. For example, if humans want to form concepts from just one example, such as learning simple character strokes, they can write a lot of text, which is typical small sample learning.
But for the machine, it is very difficult to solve this problem of “writing at a glanceâ€, because in terms of the principle of deep learning, deep learning is based on the understanding of “abstract concepts†through big-layer networks through big data. The more data there is, the better it will be, but what if there is no big data? You can only rely on small sample learning that can make a quick decision by observing a single case, or even more idealized zero-sample learning.
The article believes that Baidu artificial intelligence has achieved zero-sample learning in this breakthrough. It can be said that it is a dark horse in the field of artificial intelligence, or it will make Baidu artificial intelligence language stand out from the competition of similar products on the market such as Siri and Google Assistant.
Baidu has achieved remarkable results in the field of artificial intelligence. Its face recognition technology has been selected as one of the top ten breakthrough technologies in 2017 by MIT Technology Review. At present, the accuracy rate of Baidu face recognition has reached 99.7%. In addition, Baidu has disclosed more than 1,900 invention patents in the field of artificial intelligence, which is far ahead in China. The layout of these patents will play an important role in Baidu's preemption of technology commanding heights and planning technology development, and also set an important milestone in intellectual property protection for the arrival of the artificial intelligence era.
Baidu's “zero sample learning†project is based on its self-developed PaddlePaddle framework, or will create a true “human-machine dialogue†era. As the article said: "Baidu's new achievements in artificial intelligence will promote the advancement of natural language processing, allowing machines to one day understand the language as humans."
Difficulties in Baidu to achieve zero-sample learning in artificial intelligence language
The enhanced algorithm made by Baidu enables artificial intelligence to acquire the language organization ability of grounded air. In this regard, the US authoritative technology media Extreme Tech recently published a paper on this major advancement in the field of deep learning, Baidu artificial intelligence has successfully broken through the most difficult part of the current development of artificial intelligence: "zero sample learning."
In the article, Baidu's algorithm allows the artificial intelligence agent (Agent) in the video game to learn English, understand the written commands, and then do it through the "virtual teacher", which enables Baidu artificial intelligence to achieve zero sample learning. Learning to handle tasks without any samples is the most difficult part of the current development of artificial intelligence.
Specifically, Baidu's “zero sample learning†project allows the Agent to respond to the commands given by the teacher and then give the virtual teacher positive or negative feedback. If the agent can make the correct action, it will receive the reward, and if it fails, it will be punished. Agents can slowly understand the meaning of the language and learn to use the language correctly. Thereafter, if an unfamiliar command is encountered, the Agent can infer the correct meaning of the language to meet the expected goal, which is zero sample learning. Only the whole process takes place in a simple 2D maze, which can then be applied to the simulated 3D environment and then applied to the real scene.
Language-based use, zero-sample learning or small sample learning is not difficult for humans to understand. For example, if humans want to form concepts from just one example, such as learning simple character strokes, they can write a lot of text, which is typical small sample learning.
But for the machine, it is very difficult to solve this problem of “writing at a glanceâ€, because in terms of the principle of deep learning, deep learning is based on the understanding of “abstract concepts†through big-layer networks through big data. The more data there is, the better it will be, but what if there is no big data? You can only rely on small sample learning that can make a quick decision by observing a single case, or even more idealized zero-sample learning.
The article believes that Baidu artificial intelligence has achieved zero-sample learning in this breakthrough. It can be said that it is a dark horse in the field of artificial intelligence, or it will make Baidu artificial intelligence language stand out from the competition of similar products on the market such as Siri and Google Assistant.
Baidu has achieved remarkable results in the field of artificial intelligence. Its face recognition technology has been selected as one of the top ten breakthrough technologies in 2017 by MIT Technology Review. At present, the accuracy rate of Baidu face recognition has reached 99.7%. In addition, Baidu has disclosed more than 1,900 invention patents in the field of artificial intelligence, which is far ahead in China. The layout of these patents will play an important role in Baidu's preemption of technology commanding heights and planning technology development, and also set an important milestone in intellectual property protection for the arrival of the artificial intelligence era.
Baidu's “zero sample learning†project is based on its self-developed PaddlePaddle framework, or will create a true “human-machine dialogue†era. As the article said: "Baidu's new achievements in artificial intelligence will promote the advancement of natural language processing, allowing machines to one day understand the language as humans."
ZOOKE provides you with safe and reliable connector products, with 3.0 spacing products providing more possibilities for limited space and creating more value for the research and development and production of terminal products.
3.00 wire to board connectors,3.0 connectors,ZOOKE connectors
Zooke Connectors Co., Ltd. , https://www.zooke.com