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 AI knowledge map
Artificial intelligence AI knowledge map
In the path of stopping the investigation of artificial intelligence, we also need to understand its essence, or derive a variety of upper-level intelligence applications from its roots. In my own opinion, the learning map has played such a role.
In the past two or three years, the Internet has become so blind that it must talk about data, especially big data. One of Ma Yun’s “five new theories” is “new energy”. In the future, it is not electricity, but data. Ma Huateng said that data is one of the four major elements of AI applications. Robin Li also proposed that artificial intelligence driven by “the trinity of data, computational power, and algorithm” may become the engine for promoting economic growth and progress.
This year, the Internet giants talked about less data, and talked more about learning. Among them, Baidu, in the internal Spring Chinese speech, Li Yanhong said that these “learning figures” are all very basic components of Baidu’s artificial intelligence, and we are also The advantage associated with any company.
In addition to Baidu, Google is also actively planning on the map of learning. In May 2012, Google launched the Google Knowledge Graph and applied it to optimize search results. In 2013, Google bought natural speech processing technology company Wavii, integrating the latter technology with the Google Knowledge Graph. In 2015, Google launched the medical version of the learning map. Google Voice Search, Google Assistant, Google Lens, Google Home, and many other hardware and software products are now connected to Google Learning Maps.
Not just Baidu, Google and other search engines are planning on learning maps. Alibaba, Huawei and other players are already deploying learning maps to make themselves more knowledgeable. From Data to Knowledge, why Internet giants love this shift.
Learning about the Internet's first level of value is content or information value - just like learning, answering, encyclopedia, awareness and other learning products, he satisfies the needs of users for learning consumption, sharing, and acquisition, but in the long run The main point of learning about the value of the Internet is to play a "bridge" role from data to AI, which is indispensable and will become increasingly important.
1: Learning to use cognitive ability
Each Internet is saying that I have a huge amount of data. However, according to different types of data to stop the classification, the value of the data is not the same, a variety of data, a large amount, in order to fully play the role of these data, the invention of greater value. It is necessary to make these data become data that can be recognized and recognized by the machine, that is, knowledge-like data, so that the machine constitutes cognitive ability. Baidu AIG, the person in charge of Wang Haifeng, once proposed that AI can share perception layers and cognitive layers. Perception is a talent possessed by both humans and animals. Machines can also be stronger than humans at a certain level, but cognition is the exclusive talent of human beings. The perception of the machine has become more mature (image recognition, etc.), but there is still much room for improvement in cognitive ability. Knowledge-based data can enable the machine to form cognitive ability on the basis of perceived ability.
2: Learning allows robots to talk to humans
The current information interaction between robots and humans is anti-human - typing using input methods, etc. Even if it is simple, it requires a certain amount of learning capital, as long as touch, voice, and visual interaction are natural, with small Children use talent. Take the example between the current manager and the secretary. If there is no learning, the secretary can do as if to identify who the manager is. Such a thing, the iPhone ID's Face ID will be able to identify, but if you want to follow the manager's expression Other feelings, and make corresponding, these are the needs of knowledge to stop the analysis. For example, when you are sad, you can push a pleasant song and drink coffee to discover that there is no coffee in the coffee cup and you can automatically grind a new one, and so on. The “tired train driver identification” demonstrated by Li Yanhong at the Baidu World Conference is essentially the result of loading the knowledge map on the image recognition technology. The machine can easily understand “what performance can be called tired driving”.
3: Learning can make intelligent machine decisions
When the machine has enough knowledge to establish its cognitive ability and has its own understanding of the world, it can stop intelligent decision-making. For example, AlphaGo—related to the game of chess played by Dark Blue and so on, is different. He did not take a chess game. He stopped the player’s recognition of Go rules and the purpose of displaying the opponent’s goal. It was a real AI game. . AlphaGo Zero is an advanced version. He can stop self-learning progress without using any dance music scores and human experience. It is equivalent to constructing an autonomous learning system and becoming the exporter of learning.
Whether we are hiring a secretary or a steward, we all hope to find someone who has the idea of taking the initiative and thinking about themselves, rather than what they do. In the same way, the machine must also have certain ideas. It can make self-decisions and become intelligent. It must have enough knowledge input to make it a cognitive ability.
In the past two or three years, the Internet has become so blind that it must talk about data, especially big data. One of Ma Yun’s “five new theories” is “new energy”. In the future, it is not electricity, but data. Ma Huateng said that data is one of the four major elements of AI applications. Robin Li also proposed that artificial intelligence driven by “the trinity of data, computational power, and algorithm” may become the engine for promoting economic growth and progress.
This year, the Internet giants talked about less data, and talked more about learning. Among them, Baidu, in the internal Spring Chinese speech, Li Yanhong said that these “learning figures” are all very basic components of Baidu’s artificial intelligence, and we are also The advantage associated with any company.
In addition to Baidu, Google is also actively planning on the map of learning. In May 2012, Google launched the Google Knowledge Graph and applied it to optimize search results. In 2013, Google bought natural speech processing technology company Wavii, integrating the latter technology with the Google Knowledge Graph. In 2015, Google launched the medical version of the learning map. Google Voice Search, Google Assistant, Google Lens, Google Home, and many other hardware and software products are now connected to Google Learning Maps.
Not just Baidu, Google and other search engines are planning on learning maps. Alibaba, Huawei and other players are already deploying learning maps to make themselves more knowledgeable. From Data to Knowledge, why Internet giants love this shift.
Learning about the Internet's first level of value is content or information value - just like learning, answering, encyclopedia, awareness and other learning products, he satisfies the needs of users for learning consumption, sharing, and acquisition, but in the long run The main point of learning about the value of the Internet is to play a "bridge" role from data to AI, which is indispensable and will become increasingly important.
1: Learning to use cognitive ability
Each Internet is saying that I have a huge amount of data. However, according to different types of data to stop the classification, the value of the data is not the same, a variety of data, a large amount, in order to fully play the role of these data, the invention of greater value. It is necessary to make these data become data that can be recognized and recognized by the machine, that is, knowledge-like data, so that the machine constitutes cognitive ability. Baidu AIG, the person in charge of Wang Haifeng, once proposed that AI can share perception layers and cognitive layers. Perception is a talent possessed by both humans and animals. Machines can also be stronger than humans at a certain level, but cognition is the exclusive talent of human beings. The perception of the machine has become more mature (image recognition, etc.), but there is still much room for improvement in cognitive ability. Knowledge-based data can enable the machine to form cognitive ability on the basis of perceived ability.
2: Learning allows robots to talk to humans
The current information interaction between robots and humans is anti-human - typing using input methods, etc. Even if it is simple, it requires a certain amount of learning capital, as long as touch, voice, and visual interaction are natural, with small Children use talent. Take the example between the current manager and the secretary. If there is no learning, the secretary can do as if to identify who the manager is. Such a thing, the iPhone ID's Face ID will be able to identify, but if you want to follow the manager's expression Other feelings, and make corresponding, these are the needs of knowledge to stop the analysis. For example, when you are sad, you can push a pleasant song and drink coffee to discover that there is no coffee in the coffee cup and you can automatically grind a new one, and so on. The “tired train driver identification” demonstrated by Li Yanhong at the Baidu World Conference is essentially the result of loading the knowledge map on the image recognition technology. The machine can easily understand “what performance can be called tired driving”.
3: Learning can make intelligent machine decisions
When the machine has enough knowledge to establish its cognitive ability and has its own understanding of the world, it can stop intelligent decision-making. For example, AlphaGo—related to the game of chess played by Dark Blue and so on, is different. He did not take a chess game. He stopped the player’s recognition of Go rules and the purpose of displaying the opponent’s goal. It was a real AI game. . AlphaGo Zero is an advanced version. He can stop self-learning progress without using any dance music scores and human experience. It is equivalent to constructing an autonomous learning system and becoming the exporter of learning.
Whether we are hiring a secretary or a steward, we all hope to find someone who has the idea of taking the initiative and thinking about themselves, rather than what they do. In the same way, the machine must also have certain ideas. It can make self-decisions and become intelligent. It must have enough knowledge input to make it a cognitive ability.