The hottest Wu Enda releases AI transformation gui

  • Detail

Wu Enda released the "guide to AI transformation", decisive battle in the next 10 years

photo source: visual China

Wu Enda once said that AI will change all walks of life like electricity. The sooner enterprises start transformation, the more they will enjoy dividends. Jim Breyer, CO chairman of IDG capital, also said at the 2018 fortune global technology forum that the value of companies driven by AI in the future will exceed the current sum of Facebook, Amazon, Netflix, Google and bat. AI can connect different fields and disciplines, which will be a great opportunity to build a great value enterprise

but the problem is that not every company can hire the right executives, and not everyone has the experience to help enterprises achieve AI transformation

on December 13, Wu Enda Po issued an open letter in the media column, officially releasing AI transformation Playbook for free, translated into Chinese as a guide to AI transformation. This guide comes from the core summary of AI, which he personally led in Google and Baidu. It is also the essence of AI, which he has seen widely, heard much and consulted countless times

the following is the translation of AI transformation playbook, which is translated by qubit (id:qbitai) authorized by translator Li Shan

like electricity 100 years ago, AI technology will now change every industry one by one. From now to 2030, it will generate about $13 trillion in GDP growth. Although AI has created great value in top technology companies such as Google, Baidu, Microsoft and Facebook, the trend of value creation in the future will no longer be limited to the software field

this manual of AI transformation is based on my experience in leading Google brain and Baidu AI team. These two teams have played a crucial role in helping Google and Baidu transform into great AI companies

any enterprise may become a powerful AI company according to this manual, but these suggestions are mainly customized for large companies with a market value of 500-500 billion US dollars

I suggest that enterprises follow the following steps in the process of transformation using artificial intelligence, and I will also explain this in this manual:

1 Build up momentum by implementing pilot projects; 2. Establish an internal AI team; 3. Provide extensive artificial intelligence training; 4. Formulate AI strategy; 5. Develop internal and external communication mechanisms

1. When implementing pilot projects to prepare for the deployment of the first few AI projects, the key is to make the project succeed and not blindly pursue high-value projects

these projects must have sufficient significance, so that the initial success can help your enterprise become familiar with AI, and persuade others in the company to further invest in AI projects

their scale should not be too small to make others think they are insignificant. The key is to keep the flywheel rotating so that your AI team can gain enough momentum

the first few AI projects should have the following characteristics:

it should create cooperation opportunities for newly established products that are very important in the production of products with no more cost than before, or external AI teams (they may not know your enterprise's field) and your internal teams (they know your field very well), and develop several sets of solutions that can see results in 6 to 12 months. The project should be technically feasible. Too many companies choose projects that cannot be completed using today's AI technology, so we should let trusted AI engineers conduct due diligence on the project, and then start the project to make you more convinced of the feasibility of the project. Set a clear and quantifiable goal to create business value

when I led the Google brain team, Google was full of doubts about deep learning technology (in fact, it is the case all over the world). In order to help the team get ready, I chose Google Voice team as the first customer to greatly improve the recognition rate of google voice through close cooperation with them

speech recognition is an important project within Google, but it is not the most important. For example, its contribution to the company's profits is less than that of online search or advertising. But the voice team is more successful through deep learning technology. After it should be used for search and refueling, other teams also began to trust us, so that the Google brain team gained momentum

once other teams start to see the success of the cooperation between Google Voice team and Google brain team, we can get more internal customers. Our second largest internal customer is Google maps, which uses deep learning to improve the quality of map data. With these two successful experiences, I began to talk with the advertising team

the gradually accumulated development momentum also enables us to develop more and more successful AI projects. You can also adopt the same model in your own company

2. Set up an internal AI team

if the outsourcing partner has deep AI expertise, it can help you get ready quickly. Nevertheless, in the long run, it is more efficient to implement some projects with internal AI teams

in addition, you certainly want to keep some projects within the company in order to gain a more unique competitive advantage

if you want to build an internal team, you must obtain the approval of senior management. In the process of the rise of Internet, recruiting a CIO has become a turning point for many companies to formulate cohesive Internet use policies

in contrast, some companies have made many independent attempts, including digital marketing, data science and new stations. However, if these small-scale pilot projects cannot bring changes to other departments of the company by expanding their scale, they will not be able to make full use of the ability of interconnection

in the era of artificial intelligence, the key development momentum of many companies needs to be achieved by establishing centralized artificial intelligence teams, because such teams can help the whole company. If the professional scope is appropriate, this kind of AI team can be led by CTO, CIO or CDO (chief digital officer). Special Caio (chief AI officer) can also be arranged

The key responsibilities of the AI department are:

to establish a set of AI technology to support the whole company. In the initial stage, a series of cross functional projects are carried out to support different departments/businesses with AI projects. After completing the initial project, determine a set of repeatable processes to continue to deliver a series of valuable AI projects. Develop a consistent set of standards for recruiting and retaining employees. Develop a platform covering the whole company, which is not only helpful to all departments, but also unlikely to be developed by a single department. For example, we can consider cooperating with CTO, CIO and CDO to develop unified database standards

many companies report to the CEO through multiple business departments. After establishing a new AI department, AI talents can be allocated to different departments through matrix mode, so as to promote cross functional projects

new job descriptions and new team organizations will appear. The positions I assigned to my team members include machine learning engineer, Data Engineer, data scientist and AI product manager, which are very different from the era before AI flourished. A good AI leader can provide you with corresponding suggestions to help you determine the appropriate process

the AI talent market is now full of smoke. Unfortunately, most enterprises cannot recruit an AI doctoral student from Stanford University (even an AI undergraduate from Stanford University). After all, in the short term, the talent war is a zero sum game. Cooperating with recruitment enterprises to form an AI team may bring you many advantages

however, providing training to existing teams can also cultivate many new talents internally

3. Provide a wide range of AI training

nowadays, no company has enough AI talents inside

although the media exaggerates the salary of AI talents (the figures mentioned by the media are often outliers), it is indeed difficult to find AI talents

fortunately, various digital content channels, including coursera, ebooks and Youtube Videos, provide a very cost-effective way for many employees to receive training in new technologies such as artificial intelligence. Smart Chief Learning officers know that their job is to collect content rather than produce content, and then determine a process to ensure that employees complete the learning process

ten years ago, the so-called employee training was to hire some experts to give lectures in the office. But now it is too inefficient to do so, and the return on investment is not clear enough. In contrast, digital content costs less and gives employees more personalized experience. If you really have the money to hire experts, you should also use this face-to-face teaching method to supplement the network content

this is called "flipped classroom" teaching method. I found that if the method is appropriate, it can speed up learning and make the learning experience more enjoyable. For example, I used this teaching method when I taught deep learning courses at Stanford University

hiring several AI experts to teach in person will also help stimulate employees' enthusiasm to learn these AI technologies

AI will change all kinds of jobs. You should let everyone master the knowledge needed to adapt to new responsibilities in the era of AI. Consulting an expert can help you customize the course for your team. You can refer to the following training plan:

senior executives and senior enterprise leaders (at least 4 hours of training)

goal: let senior executives understand what AI can do for enterprises. Next, I will explain several consumable parts of universal experimental machine, start to formulate AI strategy, make appropriate resource allocation decisions, and smoothly cooperate with AI team to support valuable AI projects


1. Basically understand the business problems of AI, including basic technology, data, and what AI can and cannot do. 2. Understand the impact of AI on corporate strategy. 3. Research on the application cases of artificial intelligence in related industries

department leaders responsible for implementing AI projects (at least 12 hours of training)

Objective: department leaders should be able to determine the direction, allocate resources, monitor and track the progress of AI projects, and make corrections as needed to ensure the successful delivery of the project


1. Basically understand the business problems of AI, including basic technology, data, and what AI can and cannot do. 2. Basic understanding of AI technology, including main algorithm types and requirements. 3. Basically understand the workflow of AI project, the responsibilities of AI team and the management of AI team

AI Engineer Trainee (at least 100 hours of training)

Objective: the newly trained AI engineer should be able to collect data, train AI models, and deliver

Copyright © 2011 JIN SHI