Lei Feng Network (search "Lei Feng Network" public concern) September 1, 2016 news, Baidu World 2016 held in Beijing. Hundreds of the Assembly's chief scientist Andrew Ng interpretation of Baidu's four major brain power and depth learning platform Baidu, and Baidu announced this month will open deep learning platform.
The main speeches are as follows: 1) What current artificial intelligence can do in the end; 2) Detailed explanation of the four core parts of Baidu's brain; 3) Use of continuing learning to cooperate with third parties; 4) Reasons for using deep learning; ) Introduction to Baidu Medical Brain; 6) Artificial Intelligence is changing companies and industries.
The following is the full text of Wu Enda's speech ( Lei Feng network did not change the original meaning of the cut ): 
At the heart of Baidu's brain, we have language, images, natural language processing, and user portraits , all of which are dependent on our machine learning capabilities. Baidu is an artificial intelligence company whose products rely heavily on artificial intelligence, including search, advertising, glutinous rice, autopilot, and more. We exported Baidu's artificial intelligence technology internally to our engineers and product managers. We also arranged an artificial intelligence workshop to see what innovations our staff can use our artificial intelligence technology to achieve in a month. product. The results can be summarized in four examples:
First, we have a small team of patrol robots that use computational language processing and machine learning capabilities. They can automatically call customers to automatically tap offline risks, including accreditation, pornography, test preparation, and so on.
In the second example, we also have a small team that uses our machine learning technology as a smart writing assistant. After you write a few words, he can automatically do word replacement, word recommendation, and help you modify the article.
In the third example, when we posted the product, we got some pictures. We also need to do automatic picture tailoring. We would have got the above results, but if we use our smart tailoring capabilities, we can get the results below the picture and improve users. Experience.
In the fourth example, we have a small team to create a new product. If you take a photo and see what clothes you wear, you can automatically identify what clothes you wear and provide you with a stylish look.
Artificial intelligence is developing so fast. What can it do for you?I want to give you two ideas today.
First, if there is one thing that a normal person can do in less than a second, we may also be able to use artificial intelligence to do it automatically. If you want to enter a picture, and the output of this person's gender is good-looking, the normal person can understand this information by looking at this picture for a second, and our current artificial intelligence can also be very good.
Second, apart from working under one second, in a very short period of time, we have a second thing that can be done very well with artificial intelligence. If you can get a lot of data on a specific and repetitive event, you can get a lot of it. Data, you can use this data to predict the next result. For example, in Baidu we have a lot of advertising and related user information, seen many times, some users will not click different ads, we can use some data to predict whether users will click on ads.
I think that in a very short period of time, if a problem can be done by a normal person for less than one second, or it is a job that predicts the next result, artificial intelligence can be very effective. From a long-term perspective, I am also full of confidence in the future of artificial intelligence. I hope that in the future we will be accompanied by robots, or personalised private teaching, music composition, and robot doctors. Many of these projects are at the research stage.
The four major capabilities of Baidu's brainThe first ability of Baidu's brain is voice ability.
In our most advanced speech recognition system, we call Deep Speech, which is an end-to-end deep learning system. When you enter a speech, we can directly put it in. A huge neural network can be learned by this neural network. Output accurate results. We actually found out that if you want to do work with this kind of deep learning, you need to train very, very large neural networks. To do this, we discovered that using GPUs can train large enough neural networks to make this type of work successful. .
The most important products in Baidu already support this type of voice input, users can use these input information more conveniently, naturally, and faster to use these products. Our team is doing a new voice input method, this is a voice-based keyboard, and we also found that if you use a simple natural language processing ability to understand what you say, you can also recommend expressions, you say I am If you are in a chat and want to put in a third party friend's phone number, you can also say your needs directly. He can also automatically meet you.
The second ability is natural language processing .
Since Baidu was founded on the first day, it began as a search company. Searching required a lot of relying on natural language processing, so Baidu had very advanced natural language processing technology from the first day. You say a few words, our keyword technology can be automatically output, identify the keywords, we have a very very good translation technology, you can do automatic answer, but also understand your needs to provide you with device control.
In addition to search, natural language processing is also useful for other products. Two years ago, at the 2014 conference, I introduced to you the project of using natural language processing. That is our smart audio. If you want to listen to music at home, you sit on the sofa and play music directly to the audio. You can hear it. We think that audio may become your assistant, you can also provide you with shopping or services, but also give you useful information.
I hope that in your home, your appliances will become more and more intelligent in the future. The speakers in your home is a very, very important device. In addition to the speakers, I hope that your lighting, security systems, curtains, etc. will change in the future. Be more and more intelligent, so that it can understand your direct needs, make your home smarter, and Baidu will also start working with third-party home appliance companies to use our artificial intelligence technology to support home appliances. The company's work.
The third part is our image technology.
In fact, our image technology has been very very deep understanding of images and video, I will give you an example.
The first example, if you enter the picture on the left, we can automatically output the following sentence, telling you the contents of this picture. The second example, you enter the middle picture and ask the question, we can automatically output the answer to tell you this is a white surfboard.
What is the use of these image technologies? Very accurate face recognition ability can cooperate with third-party companies to do smart door locks, which is to use our face recognition and live recognition capabilities. Our idea is to identify your face and let you go in. The last example was done using our live recognition ability.
The fourth ability of Baidu's brain - the user's portrait.
Because we see a lot of data, we can have a very deep understanding of the user. Can know a user's gender, its age, like to eat, like to wear what clothes, like to go somewhere to travel, we Baidu also respects user privacy, so our focus, and still use this user portrait to support the internal product.
In Baidu's brain, we have various kinds of technologies. Speech images, natural language processing, and user portraits all rely on our machine learning . Our internal staff can select the best model very quickly if they want to use an already trained model, such as a speech image model, and put the data into it very quickly and easily. If a team has its own data, they sometimes want to use their own data to train their models.
In terms of machine learning, I believe that the greatest part of the value is still dependent on a class of technology called supervised learning . What does this supervised learning mean? This means that if you want to learn what you enter, you need to output something. There is a lot of user advertising information in Baidu. If you want to learn to input user advertising information, outputting a user will not click this advertisement. This is a typical example.
Use machine learning to cooperate with third partiesOver the past few years, our ability to continue learning has focused on supporting our internal Baidu work and has recently begun to use our machine learning capabilities to support the work of third parties.
Let me give you an example. Our open cloud team has worked with the Taiyuan Railway Bureau to do logistics optimization. If you have a train, how much time it takes for the train to travel from one city to another, that time is more difficult to influence, and your transit time will have a huge impact on logistics. What do you mean by transit time? If a train arrives at a train station, you transfer the goods from the train to the warehouse and how much time it takes to get from the warehouse to the truck. This is your transit time, reducing the transit time, which has a great impact on logistics.
We have collected a lot of data on logistics from the Taiyuan Railway Bureau. We put these data on the Baidu Open Cloud. We have trained several models to predict the arrival time of accurate trains and trucks, forecast future storage requirements, and future capacity requirements. Our prediction models can reduce transit time by 50%, which has a huge impact on logistics.
What is the reason for using deep learning?You all know that in Baidu, our machine learning has had a great influence on many of our products. In fact, our most advanced and best machine learning platform is our deep learning platform. Our most important products and technologies are using our deep learning capabilities. Our internal in-depth learning platform can achieve this type of success. I think there are two main reasons:
First, it is easy to learn, easy to use, and easy to use.
Second, it is highly efficient.
Because the focus of our deep learning platform is easy to learn and use, we have thousands of engineers who can learn easily and use deep learning to influence their products. Because in the Baidu deep learning platform, such a significant impact on such an important product, we think this is a very, very valuable resource for us.
I am very happy to announce that this month will open up our deep learning platform for everyone. I just said that the focus of our deep learning platform is that it is easy to learn, easy to use, and efficient. In the end, Baidu will also try its best to support the Chinese developer community. Our deep learning platform is highly efficient and supports GPU work, step-by-step training, training and learning. We hope our deep learning platform will make it easier and more convenient for your team. Effectively use the most advanced deep learning platform to change your work.
I think that the earliest use of artificial intelligence is still the Internet company, because Internet companies have huge amounts of data, and there are very good engineers. However, we have seen that artificial intelligence has begun to have a greater and greater impact on technology companies and traditional companies. For example, I have just talked about smart speakers, our smart locks, we have seen artificial intelligence has a growing impact on many traditional industries or technology industries.
Baidu medical brain introductionIn Baidu, we also have a medical brain. We are just researching and exploring whether a robotic doctor can succeed. If you use a medical brain and ask him a question, for example, what should he do if he has a fever? This medical brain will Will understand your problem, ask you a lot of questions about the situation to understand your condition, you answer his question slowly, he can also automatically identify the possibility of your condition, can also provide you with some information about the condition and Provide some suggestions.
This medical brain is certainly not a substitute for a doctor, and if the patient wants to use the medical brain information for the next step, he of course needs to discuss it with the doctor first. I hope that this kind of work in the future can provide useful information for patients and doctors. information.
In Baidu, this kind of medical brain and other work can be done faster, the main reason is our Baidu brain technology and platform. You can see that many of our most important products are using these Baidu brain technologies to succeed. Our last month's artificial intelligence workshop was also a 70-month team within a month. In a month, we can do very, very brilliant artificial intelligence innovation projects. The reason they can do it is that they can choose the most effective labor. In the intelligent technology section, they are integrated into this type of work.
Artificial intelligence is changing companies and industriesI think artificial intelligence has had a great influence on Baidu and it has changed our company. What does this have to do with you? We are opening the Baidu brain platform. I hope these technologies can help everyone change your company and change your industry.
In our Baidu brain opening website, we have written our technical capabilities. You can start to understand these technologies on your website. How do you use them? We will gradually put more training materials on artificial intelligence into This website hopes to be useful to your company and products, and it will make a big difference to us.
The theme of the Baidu World Congress this year is AI is the new electricity. A hundred years ago, electricity brought a lot of changes. Today, artificial intelligence will also bring about a big change to our industry. I hope our Baidu artificial intelligence technology can Helping you to change the industry, change society, and hope that we can help you into the era of artificial intelligence, thank you all.
The PLC splitter is used to separate or combine optical signals. A PLC is a micro-optical component based on planar lightwave circuit technology and provides a low-cost light distribution solution with the small form factor and high reliability.
Rackmount Splitter is manufactured using silica glass waveguide circuits that are aligned with a v-groove fiber array chip that uses ribbon fiber. Once everything is aligned and bonded, it is then packaged inside a miniature housing. PLC splitter has high-quality performance, such as low insertion loss, low PDL, high return loss, etc.
Advantages
Suitable for multiple operating wavelengths (1260nm – 1650nm); unstinted.
Equal splitter ratios for all branches.
Compact configuration; smaller size; small occupation space.
Good stability across all ratios.
High quality; low failure rate.
Rackmount Splitter,Fiber Optical Splitter Rackmount,1U Rack Mount PLC Splitter,PLC Fiber Optical Splitter
Ningbo Bwinners Optical Tech Co., Ltd , https://www.sijee-optical.com