learning machine
Online business empires rely on low-key methods of artificial intelligence.
This Internet business empire has chosen a low-key path in the development of artificial intelligence.
Amazon's six-page memo is very famous. Executives must write an annual report listing their business plans. Little known is that these letters must always answer a specific question: How are you going to use machine learning? Amazon managers say that answers like "not much" are not encouraged.
Amazon's six-page memo is very famous, and executives must write a page every year as required, detailing their future business plans. But what is less well known is that each letter must answer a specific question: How are you going to use machine learning? If your answer is "nothing to say", according to Amazon management, this answer is not allowed.
Machine learning is a form of artificial intelligence (ai), which mines patterns from data that can be used for prediction. When Jeff wilkie joined the company, it took root in Amazon 1999. Today, wilkie is Jeff Bezos' second-in-command. He has set up a team of scientists to study Amazon's internal processes to improve efficiency. He incorporated his researchers into the business department and turned the cycle of self-evaluation and improvement into the default mode. Soon, this cycle includes machine learning algorithms; The first one recommended books that customers might like. As Bezos' ambition grows, so does the importance of automated insight.
Machine learning is a way to realize artificial intelligence, which mainly includes specific types of data mining, and its main purpose is to predict future trends. 1999 when Jeff wilkie joined the company, the idea began to fall to the ground. Mr. Wilke is the second chairman of Amazon. He set up an artificial intelligence expert group, which is mainly responsible for the research of Amazon's internal workflow, with the aim of improving employees' work efficiency. He arranged scientists in various enterprises and institutions, fixed the continuous cycle of self-evaluation and improvement as the default mode, and soon this cycle joined the algorithm; The first generation algorithm can recommend books that customers like. With Mr. Bezos's ambition expanding, this automatic algorithm recommendation mode becomes more and more important.
However, when its technology giant partners show off
What do other tech giants have to show off?
Their artificial intelligence strength at every opportunity-Facebook's facial recognition software, Apple's Siri digital assistant or Alphabet's self-driving car and MasterGo Player-Amazon has adopted a lower-key approach to machine learning. Yes, its Alexa competes with Siri, which provides forecasting services in its cloud. But the most critical algorithm for the company's success is the algorithm it uses to continuously simplify its own operations. The feedback loop looks the same as its consumer-oriented artificial intelligence: building services, attracting customers, collecting data, and letting computers learn from these data, all of which are on a scale that human labor cannot imitate.
Technology giants seize every opportunity to show their strength in AI: Facebook has launched facial recognition software, Apple has voice assistant Siri, and Google has launched driverless and Alpha Go. Compared with these companies, Amazon has chosen a low-key path in machine learning. Alexa is an artificial intelligence service launched by Amazon, and its main competitor is Apple's Siri. Relying on Alexa's cloud platform, Amazon can provide users with forecasting services. The algorithm behind this artificial intelligence is quite distinctive. It can continuously streamline its own operation process, but the feedback loop of this AI service is similar to its client AI: a service is launched to attract target customers, collect user information, and let computers learn these data, and the scale of data processed is beyond human reach.
Mr. Potter's algorithm
Mr. Potter's algorithm
Take Amazon's fulfillment center as an example. These huge warehouses, with more than 100 in North America and more than 60 around the world, are the heart of its $207 billion online shopping business. They store and send goods sold by Amazon. In a warehouse on the outskirts of Seattle, the package shuttles on the conveyor belt at the speed of a moped. The noise was deafening-there seemed to be no one in the facility. On the contrary, in an enclosure area the size of a football field, there are thousands of yellow cuboid shelving units, each of which is 6 feet (1.8 meters) high. Amazon calls it a pod. Hundreds of robots neatly row them in and out, slide under them and drag them around. Toothpaste, books and socks are piled up in a way that seems random to human observers. However, from the point of view of the algorithm that guides this process, all this is very meaningful.
We can learn about Amazon's "execution center". They are actually more than 0/00 in big warehouse, North America and more than 60 in the world. It can be said that these warehouses are the strong heart of this company, and they have driven Amazon's online shopping trade worth $207 billion. These warehouses are used to store and distribute goods, and Amazon sells them to customers. In a warehouse in the suburbs of Seattle, the conveyor belt carries packaging supplies at the speed of a locomotive. It is difficult for you to hear any noise, and these facilities are basically fully automatic. In an area enclosed by the fence, about the size of a football field, there are some yellow square shelves, each of which is about 1.8 meters high. Amazon calls them "small warehouses". These "warehouses" are neatly arranged in a row, and hundreds of robots shuttle through them, moving them out and in. In the eyes of human beings, it is really hard to understand that toothpaste, books and socks are randomly placed on the shelves. However, under the guidance of the algorithm, this process is extremely reasonable.
Human workers, or "colleagues" in company jargon, stand in the gap of the fence around this "robot field". Some people choose items from pods brought to them by robots; Others put things into empty containers, which are quickly transported away and stored. Whenever they choose or place a product, they use a barcode reader to scan the product and related shelves, so that the software can track it.
Human employees, or "human partners" as Amazon calls them, mainly provide auxiliary services for robots. Their workplace is located on the platform between fences, and the inside of the fence is the so-called "robot area". Robots keep carrying small warehouses, some employees take goods from them, and some put them back into empty warehouses. But whether employees take it out or put it back, they will scan the goods and the corresponding shelves with bar code scanners, so that the software system can record the running path of the goods.
The person responsible for developing these algorithms is Amazon's chief robotics expert Brad Porter. His team is Mr. wilkie's fulfillment center optimization team. Mr. Porter pays attention to the "pod gap", that is, the time that human workers have to wait before robots drag pod to their workstations. Fewer and shorter intervals mean less downtime for human workers, faster flow of goods through the warehouse, and finally faster Amazon delivery. Mr. Potter's team keeps trying new optimizations, but be careful when promoting them. The traffic jam in the robot field is simply hell.
Brad Porter is the main developer and manager behind these algorithms, and also the chief robotics scientist of Amazon. The team he formed is an optimized version of Mr. Wilke's team, and its main service object is the execution center. Mr. Porter mainly focuses on how to narrow the gap between small warehouses and how to reduce the time for human employees to wait for robots to deliver goods on their platforms. For human employees, smaller and smaller customs clearance means shorter loading and unloading time, faster cargo transportation process and faster delivery service. All along, Mr. Porter's team has been trying new optimization strategies, but every promotion is very cautious, because the traffic jam in the "robot area" is a very serious and terrible problem.
Amazon Web Services (aws) is another part of the core infrastructure. It supports Amazon's $26 billion cloud computing business, which allows companies to host websites and applications without their own servers.
Amazon Web Services (AWS) is another component of its core infrastructure. Its existence maintains Amazon's cloud computing business worth $260 billion. Using this network system, companies can open their own websites or develop their own applications without servers.
The main use of aws for machine learning is to predict computing requirements. When Internet users flock to customer service, insufficient computing power will lead to errors and sales losses, because users will encounter wrong pages. "We can't say we are out of stock," said Andy Jassy, the boss of aws. To make sure they never need it, Mr. Jassy's team processes customer data. Amazon can't see what is hosted on its server, but it can monitor how much traffic each customer gets, how long the connection lasts, and how stable the connection is. As in its fulfillment center, these metadata provide information for machine learning models, which predict when and where aws will see requirements.
The main use of AWS in machine learning is to predict computing requirements. When Internet users flood into the client, the lack of computing power will lead to many mistakes, such as the user entering the wrong page and the transaction has to be cancelled. "It can't be said that there is no inventory." Andy Jassy, the boss of AWS, said that in order to ensure that this network system will never go wrong, his team collected and analyzed a lot of customer data. Although Amazon can't know the content on the server, it can detect how much traffic customers get, how long their connection with the server lasts, and the quality of this connection. In Amazon's execution center, the machine learning model relies on the input of these metadata and then runs. The main function of these models is to predict when and where the AWS system may generate computing requirements.
One of aws's biggest customers is Amazon itself. One of the main things that other Amazon companies want is prediction. The demand is so high that aws has designed a new chip called Inferentia to handle these tasks. Mr. Jassy said that reasoning would save
Amazon invests in all the machine learning tasks it needs to run to keep running normally and attract customers to use its cloud services. He said: "We believe that this can at least increase the cost and efficiency by an order of magnitude." . Alexa's algorithm for recognizing sounds and understanding human language will be a big beneficiary.
One of AWS's biggest customers is Amazon itself. At the same time, the demand for AWS in other Amazon businesses is also concentrated on its forecasting ability. Because of the huge amount of calculation, researchers have designed a new chip for AWS to handle these tasks, which is called reasoning. Mr. Jesse said that this chip will save Amazon a lot of money on various tasks of machine learning, and at the same time attract more customers to choose its cloud service. Mr. Jesse also said that "reasoning will bring an order of magnitude improvement to the cost efficiency of the company." Alexa, who can recognize sounds and understand human language, will bring endless benefits to its own algorithm development.
The company's latest algorithmic adventure is Amazon Go, a cashier-free grocery store. Hundreds of cameras observe shoppers from above, converting visual data into 3d contours to track their hands and arms when handling products. The system can see which goods shoppers have selected and put them in their Amazon account when they leave the store. Dilip Kumar, the boss of Amazon Go, stressed that the system is tracking the physical movements of shoppers. He said that it did not use facial recognition to identify them and link them with their Amazon accounts. Instead, this is done by brushing the bar code at the door. The system attributes the subsequent actions of the 3d profile to the swiped Amazon account. This is an advantage of machine learning, which processes data from hundreds of cameras to determine what shoppers have bought. No matter how hard he tries, your reporter can't fool the system and steal something.
In terms of algorithm exploration, the latest achievement of this company is Amazon Go, a grocery store without cashiers. Hundreds of cameras in the store monitor customers' behavior from above all the time, and convert the collected visual data into three-dimensional user information. The purpose of these data is to track the arm movements of customers when they take goods. In this way, the algorithm system can know which goods the customer has taken, and automatically send the bills of these goods to the customer's Amazon account when the customer leaves the store. Dilip Kumar, the boss in charge of Amazon Go project, stressed that the purpose of this system is to track customers' body movements, and did not use facial recognition to identify customers' information to connect their Amazon accounts. This system is the "ode" of machine learning. It collects information from hundreds of cameras to determine what customers have taken. Maybe you are going to steal a product, but these camera systems will not be fooled easily.
Meet the purpose
Cut clothes/coats/clothes/skirts according to the drawing ―― as the case may be.
Artificial intelligence body tracking also appeared in the fulfillment center. The company has a pilot project, which is internally called "Nike Intention Detection" system. It does what Amazon Go does for shoppers: it tracks what they choose and put on the shelves. The idea is to get rid of hand-held barcode readers. This kind of manual scanning takes time and is a trouble for workers. Ideally, they can put items on any shelf they like, and the system will observe and track them. As always, the goal is efficiency and maximizing the speed of product flow. "For employees, it feels very natural," Mr. Porter said.
Artificial intelligence motion tracking is also very useful in the execution center. Amazon has launched a pilot project, which is called "Nike Intention Detection" system within the company. Its operation principle in the execution center is the same as Amazon Go: tracking the trajectory of goods taken out and put back on the shelf. The idea is mainly to eliminate the previous hand-held barcode scanner, because such input work is a waste of employees' time and is also very troublesome to operate. Ideally, employees can put goods on any shelf under the monitoring and tracking of the system. Amazon's goal is always to improve efficiency and maximize the circulation rate of products. In Mr. Potter's words, "all our human employees feel that this process is very natural."
Amazon's cautious data collection method protects it from the government censorship that Facebook and Google have recently faced. Amazon collects and processes customer data for the sole purpose of improving the experience of its customers. It does not operate in the gray area between satisfying users and customers. The two are often very different: people get social media or search for free because advertisers pay Facebook and Google to get users. For Amazon, they are basically the same thing (although it is playing with advertising sales). Regulators are indeed worried about Amazon's dominant position in its core business of online shopping and cloud computing. This ability is based on machine learning. It shows no sign of abating.
In terms of data collection, Amazon has chosen a very cautious path. Therefore, compared with Facebook and Google, the relevant government departments have much less censorship of Amazon, and some parts can even be exempted. The main reason is that the user information collected and processed by Amazon is only used to improve the user's operating experience, and there is no gray area between meeting the needs of users and consumers. The difference between data users and manufacturers (consumers) is usually obvious: people can use social media or free search engines because advertisers can reach consumers by paying advertising fees to Google and Amazon. For Amazon, the two are basically the same person (although Amazon doesn't care much about advertising revenue). But Amazon also faces some regulatory issues, such as its monopoly position in online shopping and cloud computing. However, the establishment of this status is based on strong machine learning, and there is no sign that they are declining.