1, Youtube automatic subtitles
In 2009, Google used the existing speech recognition technology to add subtitles to Youtube videos, so that people can watch movies from various countries without interruption.
2, artificial intelligence bionic eye
The Royal Eye Hospital in Manchester, England successfully completed the world's first artificial intelligence bionic eye transplant operation. This bionic eye device, named ArgusII, consists of two parts: internal implantation and external wearing. Therefore, the Second Vision Company has developed an artificial intelligence eyeball, which can help the blind to restore their vision and have a pair of bright eyes.
3. News writing robot
Associated Press, the automatic writing system developed by Wordsmith platform can help enterprises save manpower and allow journalists to do more meaningful work, such as news planning and news source development.
4.Skype real-time translation tool
Microsoft combines speech recognition technology with static machine translation technology to develop a Skype real-time translation tool, which can help people solve the language barrier problem. Today, Skype real-time translation tool has supported language translation in more than 50 countries, such as French and Japanese.
5、IBMWatson
IBM released WatsonAnalyTIcs. WatsonAnalyTIcs realizes the cognitive service based on natural language, which can provide real-time forecasting and visual analysis tools for business people. WatsonAnalyTIcs will launch a free value-added application version based on cloud services at the end of this year, which can be used on computers and mobile devices.
WatsonAnalytics can provide self-service analysis functions, including data access, data cleaning and data warehouse, which can help enterprise users to obtain and prepare data, analyze and visualize the results on this basis, and provide basis and convenience for users to take effective actions and further interaction.
6. Google artificial intelligence project DeepMind
DeepMind, Google's research and development department in London, has developed artificial intelligence technology that can play video games independently. The computer system based on DeepMind technology can learn at an amazing speed, quickly master the gameplay and master the winning method of the game. Previously, the team called it a deep Q-network learning network. As long as we observe the changes of the game screen and game scores, and analyze the gameplay and algorithms, we can get "customs clearance skills" and get high scores, which can reach the level of professional human players.
At present, the system has passed 49 game tests with the same algorithm, network architecture and parameter settings, and has been able to master 22 games (including the above-mentioned Space Invaders), reaching the level of expert games. This system further proves that artificial intelligence can master game skills through deep learning, gain the same control as human beings, and even surpass human beings in some aspects.
Ten application fields of artificial intelligence:
1 agriculture
Many artificial intelligence technologies have been applied to agriculture, such as spraying pesticides and weeding by drones, real-time monitoring of crop conditions, material procurement, data collection, irrigation, harvesting and sales. Through the application of artificial intelligence equipment terminal, the output of agriculture and animal husbandry has been greatly improved, and many labor costs and time costs have been greatly reduced.
Step 2 communicate
Intelligent call system, customer data processing (order management system), communication troubleshooting, virus interception (360), etc. ), harassment information interception, etc.
3. Medical treatment
Using the most advanced information technology of Internet of Things, we can realize the interaction between patients and medical staff, medical service institutions and medical equipment, and gradually develop and realize enterprise informatization. E.g., smart wearable devices for health monitoring), and an intelligent medication system that automatically prompts medication time, contraindications and remaining dosage.
4. Social security
Security monitoring (data real-time networking, public security system real-time investigation and analysis data), telecom fraud data locking, criminal arrest, fire rescue field (fire fighting, personnel rescue, special area operation), etc.
5. Transport field
Route planning, driverless cars, speeding, illegal driving and other behaviors.
6. Service industry
Catering industry (ordering food, delivering food, recycling tableware, cleaning), etc. And reservation system (hotel, air ticket, air ticket, etc. ) inquiry, reservation, modification, reminder, etc.
7. Financial industry
Big data stock analysis, securities, industry trend analysis, investment risk estimation, etc.
8. Big data analysis
Weather query, map navigation, data query, information technology promotion and recommendation engines actively discover the current or potential needs of enterprise users and actively push information to users' browsing pages through research, analysis and processing of control algorithms based on the data generated by network users' behaviors and users' browsing behaviors.
9. Computer vision
Machine vision plays an important role in many situations that human vision cannot perceive, such as accurate legal perception, dangerous scene perception, invisible object perception and so on. Machine vision highlights its advantages. At present, machine vision has been applied to parts identification and positioning, product inspection, mobile robot navigation, remote sensing image analysis, monitoring and tracking, national defense system and other fields.
10, intelligent control
Intelligent control refers to the technology that drives intelligent machines through independent innovation to achieve internal control management goals without human intervention. The development of control theory has a history of 100 years. It has experienced the development stages of classical control theory and modern control theory, and entered the development stage of large-scale system theory and intelligent control theory.
Domestic products belonging to artificial intelligence are
The artificial intelligence products belonging to the family include smart home system, smart door lock, wearable device, smart speaker, smart TV, smart light bulb and so on.
What are the application fields of artificial intelligence?
application area
Machine translation, intelligent control, expert system, robotics, language and image understanding, gene programming robot factory, automatic programming, aerospace application, huge information processing, storage and management, performing complex or large-scale tasks that cannot be performed by combined organisms, etc.
It is worth mentioning that machine translation is an important branch of artificial intelligence and the first application field. However, as far as the existing machine translation is concerned, the translation quality of the machine translation system is far from the ultimate goal; The quality of machine translation is the key to the success of machine translation system. Professor Zhou Haizhong, a mathematician and linguist in China, once pointed out in the article "Fifty Years of Machine Translation": To improve the quality of machine translation, the first thing to be solved is the language itself rather than the programming problem; It is certainly impossible to improve the quality of machine translation by relying on several programs to make a machine translation system; In addition, it is impossible for machine translation to reach the level of "faithfulness and elegance" before human beings have figured out how the brain makes fuzzy recognition and logical judgment on language. After smart home, artificial intelligence has become a new trend in the home appliance industry, and Changhong is becoming the first home appliance giant to set off this wave. Changhong released two new CHiQ smart TV products, focusing on mobile phone remote control, take-away, watch at any time, watch in categories and other functions.
"Core" Track, "Small" System, Healthy New Ecology of Huami Wearable Equipment
Today, health has become everyone's concern, and wearable devices have begun to influence and change people's lives with meticulous companionship!
Goerge smoot, winner of the Nobel Prize in Physics, once participated in a project that focused on the clinical application of wearable devices and artificial intelligence technology. With the help of smart wearable devices, automatic early warning of acute myocardial infarction can save lives to the greatest extent.
He believes that if wearable devices are combined with artificial intelligence, through the popularization of "wearable devices+artificial intelligence", people can know more about their bodies, really diagnose and treat diseases at an early stage, and even improve human life expectancy 10 years old.
Dreams follow reality. Today, with the rapid development of technology and intelligence, various wearable devices emerge one after another, and a new wave of wearable devices is blowing up in the domestic and foreign markets. This trend is catching up with the increasingly saturated smartphones, opening a new path of "technology changes life".
At present, smart wearable devices on the market can be used to track and analyze personal exercise health data such as heart rate, sleep and blood oxygen, and help people complete health monitoring and early warning of heart-related diseases without leaving home.
In the view of Huami Technology, it is far more than that. More importantly, Huami believes that health is the core value of the smart wear industry, and even the core value of the entire Internet of Things intelligent hardware industry.
To provide the best products with excellent functions, extreme experience and smooth use, it is not enough to use general-purpose chips and magic-changed systems alone. In other words, magic is simple and rude, and self-research and customization require a lot of investment. Only in this way can we make the ultimate products that meet the needs of users.
Is that enough? "Not satisfied" Huami is still seeking change! On the basis of the health function of Huami products, the PumpBeats blood pressure monitoring engine, which has been developed for five years, shows Huami's scientific and technological research ability. The engine has completed the first phase of the experiment, and the wearable device equipped with the engine can measure blood pressure for 30 seconds with one button.
In addition, Huami hopes to "let nuclear magnetic equipment enter thousands of households like PCs in those days" and promote the research and development of global portable nuclear magnetic resonance MRI technology through investment. "Small" and cleverness have always been Huami's specialty. Portable magnetic resonance equipment with a height of only 1.5m and a weight of less than 0.8t inherits the lightweight characteristics similar to wearable devices.
Behind all efforts, Huami has continuously passed on the concept of "connecting science and technology with health", continuously enriched and promoted the "core cloud" strategy, and passed on thoughts on the future of the health industry.
The change of people's healthy life has just begun!
"core" orbit, "small" system
In goerge smoot's vision, in the future, wearable devices will enter all aspects of life and even be printed directly on the skin. As long as you are willing and accepted, you can be completely managed.
According to the latest data from IDC, the global smart wearable devices have broad prospects. 20 16 ~ The compound growth rate of total shipments of wearable devices is 44.5%. In the next five years, the compound growth rate of global smart wearable device shipments will be about 25%, and the estimated shipments in 2025 will be 65.438+358 million units, maintaining a strong growth trend.
This is a blue ocean of technology and intelligence, and it is also a hot land full of imagination. All kinds of enterprises are one after another, noisy and lively! However, counting the products on the market, many smart wearable devices use general-purpose chips and magic change systems, which cannot meet the needs of users and also limit the imagination of smart wearable devices.
Wang Huang, founder, chairman and CEO of Huami Technology, has a deep understanding of this. He found that to achieve the ultimate user experience of a smart watch and smart bracelet, it is almost impossible to find a suitable chip and operating system in the market, integrate it into the square inch, and make those functions that users want.
It's not that I haven't thought about shortcuts, but shortcuts are not Huami's way, because Wang Huang refuses to compromise on product experience. Wang Huang said that the extreme of hardware is the chip, the extreme of software is to make your own operating system, and the extreme of algorithm is to make a set of artificial intelligence algorithms. When these three technologies are developed to the extreme, the final product is the ultimate product.
Therefore, there are thousands of roads, and there is only one self-research for Huami. Only through self-research can we meet the needs of users and complete the products with the best performance, the lowest cost and the best experience.
"Those who are serious about hardware must make their own software", "Those who are serious about big data must make their own sensors" and "If you want to complete beautiful algorithms, you must make chips, otherwise, you can't find a chip that fully meets your algorithm in the industry", which is the obsession of Huami and Wang Huang.
Since the first generation of 20 14 was produced in mi band, the shipments of self-owned brands Amazift and Zepp smart watches have ranked among the top four in the world. Huami constantly integrates hardware, software and algorithms to make products more user-friendly and more suitable for wearable devices.
Obsession can make people stubborn, and it can also make generation show great power. Obviously, Huami belongs to the latter. In 20 18, Huami released its own wearable device chips-Huangshan 1 and Huangshan 2. At the recent NextBeat202 1 conference, Huami once again released the same series of new generation chips-Huangshan 2S.
This Huangshan 2S adopts dual-core RISC-V architecture, which not only integrates GPU and small core, but also makes small core power consumption lead the world. Huangshan 2S has higher computing efficiency and lower power consumption. NPU with convolutional neural network acceleration technology greatly improves the computing performance of local AI data. The speed of identifying atrial fibrillation is 7 times that of Huangshan 1 and 26 times faster than that of pure software algorithm.
Compared with most instruction sets, RISC-V architecture has the characteristics of open source, simple architecture and easy transplantation of related systems. Three consecutive generations of chips have adopted RISC-V architecture, and Huami's idea of creating ecology is obvious. Corresponding to this idea, at this conference, Huami also released the self-developed smart watch operating system ZeppOS.
ZeppOS is characterized by lightness, low power consumption and openness. ZeppOS system package is only 55MB, which is about 1/28 of Apple watchOS. ZeppOS system not only ensures the portability and endurance of the system, but also maintains smooth operation, and the frame rate of most internal interfaces can exceed 60fps.
Low-power self-developed chips and lightweight and agile "small" systems constitute Huami's "one machine and two wings", which greatly reduces Huami's dependence on other chip manufacturers. At the same time, through the self-developed system, Huami has more imagination space.
And its operating system not only supports Huangshan 2S, but also supports other third-party smart watches and even the processors of the whole RISC-V ecosystem. As simple and easy as possible, intelligent interaction is easily realized, and it is natural for others to use it.
"Openness is the gene of Huami," Wang Huang said.
Wearable health password
Looking back at the wearable devices and algorithms released by Huami in the past, health and exercise have become its main product features.
In the field of health, wearable devices can perceive, record, analyze and manage relevant data with the support of relevant software, so that users can perceive their physical condition more intuitively. Huami focuses on health, closely follows the actual needs and characteristics of wearable products, and practices the mission of "connecting health with technology".
Prior to this, Huami's heart rate engine monitored arrhythmia and atrial fibrillation 24 hours a day. Whether it's a heart attack or an irregular heartbeat caused by staying up late, any subtle signal will be caught.
Moreover, Huami innovatively applied blood oxygen detection to sleep respiratory syndrome. Once respiratory obstruction is detected and blood oxygen drops rapidly, Huami's smart watch will immediately vibrate to remind people to help.
But man is a complex life, and there are many problems that technology and intelligence can't solve temporarily, such as blood pressure. At present, hypertension has become a common and frequently-occurring disease in the world, and it is also the most important risk factor for cardiovascular and cerebrovascular diseases. Data show that the number of patients with hypertension in China has reached 245 million, and the number of deaths related to hypertension is as high as 2 million every year.
Monitoring blood pressure by scientific means is much more complicated than heart rate. Because heart rate measurement is a kind of frequency measurement, as long as the frequency of pulse wave beating is monitored, the structural information, morphological information and related information of pulse wave itself are not very concerned.
Blood pressure is different. Blood pressure needs to monitor the information of pulse wave fluctuation, not only in frequency, but also in morphology. It is much more complicated to monitor and obtain relevant information with optical sensors, and it will also be affected by individual differences such as skin, hair and sitting posture.
Based on the third generation sensor and blood pressure research algorithm, Huami has continuously collected the data of hypertension patients and improved them, and successfully developed the PumpBeats blood pressure monitoring engine. Smart watches equipped with this engine can effectively capture blood pressure signals and waveforms, and measure blood pressure with one button in 30 seconds, which is an important step in the field of non-invasive and sleeveless blood pressure measurement technology.
According to the clinical experimental data of Huami Technology and Peking University First Hospital, compared with the measured values of standard electronic sphygmomanometer, the average absolute deviation of PumpBeats blood pressure monitoring engine is less than 5. 14 mm Hg, and the diastolic pressure is less than 4.88 mm Hg, which fully verifies the reliability of this technology.
In the future, Huami will continue to promote 24-hour continuous blood pressure monitoring to achieve the perfect combination of night sleep monitoring and passive blood pressure monitoring, and hope to explore a new way to screen occult hypertension through long-term monitoring of smart watches.
Active health "disease prevention" is the focus of Huami. Through technical research, more and more ordinary people can get their own body data more easily, which will be combined with the data of wearable devices and finally integrated into health data on personal cloud. This is Huami's wearable health password.
Based on the strength of Huami's own sensors, chips and algorithms, it can realize richer health monitoring functions and bring more dimensions of health data to users. By the end of 2020, Huami smart wearable devices had recorded more than 1, 5 1 trillion steps, slept128 million nights, and recorded heart rate120.8 million hours.
With more and more data, the competition barrier will be higher and higher, the algorithm will be more and more accurate, and Huami's knowledge of users' health will be deeper and deeper.
Wang Huang said that the underlying logic of particularly valuable apps is basic necessities. Another is health, which is one of the lowest needs of people. In the field of health, Huami will produce core competitiveness.
It is precisely because we have firmly grasped the real needs of users that Huami's shipments and rankings are in the forefront. According to the data, as of March 3 1 and March 202 1, the global total shipments of Huami smart health equipment exceeded10.70 billion units. In the first quarter of this year, the shipments of Huami's own brands Amazfit and Zepp smart watches have ranked among the top four in the world.
On the road of connecting technology with health, technology has helped users' health, and health has made Huami's wearable devices.
Create a healthy closed-loop ecology
Any achievement is not inevitable, but accumulated from the moment you decide to do it.
According to the data released by Huami, in the past three years, Huami has invested more than 400 million yuan in R&D every year, and last year's investment reached 538 million yuan. If compared with other Internet of Things companies in the same period, the investment has doubled.
High R&D investment has made Huami's patent application in the field of big health 1 000. Among them, invention patents account for half, and the United States has nearly 100 applications for core invention patents. From the perspective of overall competitiveness and commercial value, Huami has become one of the wearable companies with the strongest shipments in the world.
Huami, who is good at small and fine, is different from giants such as Apple after all. Wang Huang once compared Huami with Apple, and thought that "Apple relies on brand, and we rely on technology and cost supply chain control".
At the product level, in addition to smart bracelets and smart watches, Huami has also launched related peripheral products such as treadmills and running shoes, which can monitor users' health data in real time through multiple channels and dimensions. At the algorithmic level, Huami also launched its own sports engine, personal health assessment system, heart rate engine, blood oxygen engine, sleep engine and so on.
Now, based on self-developed chips, operating systems, sensors and artificial intelligence algorithms, Huami has objectively started the vertical integration of related industries. When users realize the value of health, they also agree with Huami's products. From a systematic point of view, Huami has been laying out the industrial chain to create a closed-loop healthy ecology.
Moreover, Huami has cooperated with the team of Academician Zhong Nanshan, PAIHealth, a medical and health software technology company, Royal DSM Group of the Netherlands and Prudential Asia of the United Kingdom. Take smart watches as the entrance, build a powerful human health data analysis engine, build a healthy big data platform, and build a global healthy ecology.
It is worth noting that Huami's health products and services are not limited to C-end, but also oriented to B-end. For example, investment will promote the research and development of global portable nuclear magnetic resonance technology, thus opening the era of portable nuclear magnetic resonance, and it is expected to enter various medical institutions and hospital clinics. Therefore, it is easier for ordinary people to obtain their own body data, which will be combined with wearable devices in the future, so that everyone can have a clearer understanding of their health.
From a long-term perspective, Huami's ideal ecology is to connect health through various wearable devices under the same system and facilitate all aspects of people's lives.
The progress of technology will not stop. Huami started with hardware, but always took people as the center, built a soft and hard integration strategy of "product+service" and entered the field of big health service step by step. This is Huami's ecological dream.
abstract
The vitality of any form of wearable device depends on the capabilities it brings to users. Therefore, in order to build a better wearable device, enterprises must make it have enough effective functions, but it can also give users a simple experience in complexity.
How to make smart wearable devices play the ultimate role and change people's future lives in the ultimate state is the challenge of Huami and Wang Huang's insistence.
The past is the order, and everything starts with action and longing!
AI Foundation | What are the infrastructures?
We know that infrastructure is the original foundation for the existence of artificial intelligence products, so what are the infrastructures?
A sensor is a physical device or biological organ, which can detect or feel external signals, physical conditions or chemical components, and give the detected information to other devices or organs. For example, people's skin can feel cold, hot, humid and dry. Sensors transmit these signals to the brain, and the brain then instructs people to do actions such as adding clothes, reducing clothes, drinking water and opening windows for ventilation.
The function of the sensor is to convert one signal mode into another. How to classify sensors?
According to different application fields, sensors are divided into the following categories: pressure sensor, temperature sensor, PH sensor, flow sensor, liquid level sensor, ultrasonic sensor, immersion sensor, illumination sensor and so on. There are many kinds of sensors, and mainstream sensors can be divided into the following categories:
(1) biosensor
It is an analytical device that converts various types of biological reactions into electrical signals. At present, biosensors are mainly used in medical treatment, food detection, river surrounding detection and other fields.
(2) Photosensitive sensor
It is a sensor that converts optical signals into electrical signals, which can be understood as simulating human visual ability. Image sensing CCD, CMOS, human body induction lamp, human body sensing switch, light-controlled switch and mobile phone screen brightness level are all application examples of photosensitive sensors.
(3) Sound sensor
The sound sensor can be understood as the ear of human AI products. Sound sensors are used in common corridor sound control lights.
(4) Chemical sensors
It is sensitive to various chemicals and converts its concentration into electrical signals, which is the "nose" of AI products. At present, chemical sensors are widely used in air pollution monitoring, mineral resources detection, meteorological observation, industrial automation, agricultural preservation and other fields.
Generally speaking, sensors are mainly used in four kinds of artificial intelligence products, namely, wearable applications, advanced driver assistance systems, health monitoring and industrial control.
With the application value of deep learning such as image recognition, speech recognition and search/recommendation engine widely recognized, the two key links of this process-training and reasoning need strong computing power. Therefore, the chip has become the key to establish a competitive barrier in the AI field.
What are the categories of AI chips? According to the purpose, it can be divided into the following three categories: simulation training, cloud reasoning and device-side reasoning.
(1) simulation training link chip
This process needs GPU to improve the training efficiency of depth model, because it has to deal with massive data and complex depth neural network. Compared with CPU, GPU has powerful parallel computing ability and floating-point ability, and can also provide faster processing speed, less server investment and lower power consumption. In addition to PGU, TPU provided by Google can also provide deep network acceleration in training.
(2) Chip inferred from the cloud
At present, mainstream AI applications need to provide services through the cloud, transmit the collected data to the cloud server, then infer tasks from the server, CPU, GPU and TOPU, and then return the processing results to the terminal. Therefore, it is to put the reasoning link in the cloud.
(3) terminal equipment chip.
It can also be called the chip of embedded devices, such as smart phones, smart security cameras, robots and other devices.
According to the degree of customization, it can be divided into general chip, semi-customized chip (FPGA chip) and fully customized chip (ASIC).
3. Basic platform
(1) big data technology
Big data technology is the premise of artificial intelligence, and there is only one goal of big data-mining value from massive data.
(2) Cloud computing technology
According to the definition of National Institute of Standards and Technology, cloud computing is a pay-per-use model, which provides available, convenient and on-demand network access, enters a configurable computing resource pool (including network, server, storage, application software and services), and can be provided quickly with little management work or little interaction with services.
Cloud computing technology greatly reduces the economic consumption of enterprises.