Heart of Machine reporting, participating: Nan Ze, Qian Zhang, Yazhou Li.

How can ceos transform their businesses with AI? Ng tweeted in August that he would release a presentation to the company’s management on the transformation of the AI industry after talking to ceos. This time, Ng, a Stanford professor and founder of Coursera, an online course, plans to teach managers the lessons of “All in AI” as an educator.



Ng’s AI Transformation Guide has just come out. As you prepare for the AGE of AI, do you need to know something about it?

PDF download address: d6hi0znd7umn4.cloudfront.net/content/upl…

“The AI Transformation guide incorporates many of the lessons I learned while building AI research teams at Google and Baidu, as well as conversations I have had with ceos of other companies, including many outside the tech industry.” “Ng told VentureBeat in an interview ahead of the guide’s release.

Ng believes that managers trying to transform their companies into AI-driven enterprises are facing some challenges and may make some common mistakes. Focusing only on data and engineering, or misestimating the role of AI, could lead to failure, he warns.

Let’s take a look at Ng’s AI transformation guide:

Ai will revolutionize every industry, just as electricity did 100 years ago. It will add about $13 trillion to GDP between now and 2030. At the same time, AI has created tremendous value for tech giants like Google, Baidu, Microsoft and Facebook, and most of the added value it will create will go beyond the software industry.

The AI Transformation Guide glean insights from the development of Google Brain and Baidu AI team, which have played an important role in the AI transformation of Google and Baidu. By following this guide, it is possible for any enterprise to become a strong AI company, although the recommendations are tailored for large companies with market capitalization of between $500 million and $500 billion.

Here are my recommendations for the transformation of enterprise AI, which are explained in detail in the Guide:

1. Pilot projects gain momentum

2. Build an internal AI team

3. Provide extensive AI training

4. Plan the right AI strategy

5. Establish internal and external communication

1. Pilot projects gain momentum

The success of the first AI project is more important than the most valuable AI project. These AI projects need to be meaningful, because early success will help familiarize your company with AI and convince the rest of the company to invest in AI projects. In addition, these projects should not be too small to be considered unimportant. The important thing is to get the wheels turning and the AI team motivated.

Some suggestions for the first AI projects:

  • For a new AI team or an external AI team that doesn’t know your business well enough, these projects need to be able to work with an internal team that knows your business well enough to build AI solutions that begin to gain traction within 6-12 months.

  • These projects need to be technically feasible. Many companies are starting projects that would not be possible with today’s AI technology. Having AI engineers do due diligence before starting can ensure the viability of these AI projects.

  • Clearly define and measure the business value that the project can create.

When I led the Google Brain team, deep learning was viewed with great skepticism at Google (and around the world more broadly). To help power Google Brain, I chose the Google Voice team as my first internal customer and worked closely with them to improve the accuracy of Google speech Recognition. Speech recognition is a meaningful project for Google, but not the most important. Applying AI to web search or advertising is more important. But through our success with speech recognition, other teams began to trust us, and Google Brain gained momentum.

Once other teams start seeing Google Brain’s success in speech recognition, we’ll be able to get more internal customers. Google Brain’s second big internal customer is Google Maps, which uses deep learning to improve the quality of its map data. With these two results in mind, I started talking to the advertising team. Motivation gradually leads to more and more success. This process can be replicated within your company.

2. Build an internal AI team

While working with external senior AI experts can help you gain initial momentum quickly, it’s more efficient in the long run to have an internal AI team to execute some projects. In addition, you’ll want to do projects in-house to build a competitive advantage.

It’s important to build an internal team and hire people at the executive level. In the rise of the Internet, for many companies, hiring a CIO is very important for the company to integrate the Internet strategy. By contrast, companies that do individual experiments, from digital marketplaces to data science experiments to launching new websites, struggle to take advantage of the Internet’s power because these small experiments are hard to scale and transform.

For many companies in the AI space, a key moment is to build an AI-focused team that can help the entire company. With the right functions, such a team could be led by a CTO, CIO or CDO (chief data officer) or an assiduous CAIO (chief AI officer). Their key responsibilities include:

  • Create the required AI power for the entire company.

  • Execute a series of cross-functional projects to support different departments/businesses with AI projects. After completing the initial project, establish a repetitive process to continuously deliver a series of valuable AI projects.

  • Develop consistent hiring and retention standards.

  • Develop platforms for the entire company that are useful to multiple departments/business groups that could not be developed by a single department. For example, consider working with Ctos/CIOs/Cdos to develop uniform data warehousing standards.

Many companies have multiple business units reporting to the CEO. With a new AI team, you’ll be able to pool AI talent across different departments to drive cross-functional projects.

A new job description and a new team organization will emerge. I now organize my team’s work in different ways than in the pre-AI era, with the roles of machine learning engineer, data engineer, data scientist, and AI product manager. A good AI leader will help you set up the right process. The war for AI talent is on, and unfortunately, most companies will have a hard time hiring Stanford AI Ph.D.s, or even Stanford AI undergraduates. The war for talent is essentially a zero-sum game in the short term, so working with a recruiting partner who can help you build your AI team can be a big advantage. However, training your existing team is also a great way to develop a lot of new talent internally.

3. Provide extensive AI training

No company has enough AI in-house talent. While media reports of AI salaries are somewhat exaggerated (the numbers cited are often outliers), AI talent is in short supply. Fortunately, with the growth of digital content — including online courses such as Coursera, e-books and YouTube videos — it is more cost-effective than ever to train large numbers of employees in new skills such as AI. Smart Clos (chief learning officers) know their job is to plan, not create content, and then set up processes to ensure employees complete the learning process.

A decade ago, employee training meant bringing consultants to the company for classes. But it’s not very efficient, and the ROI isn’t clear. Digital content, by contrast, costs less and gives employees a more personalized experience. If you can afford to hire consultants, what they teach should complement what is available online. (This is called the flipped classroom. I’ve found that when done well, it can speed up learning and make for a more comfortable learning experience. My intramural deep learning course at Stanford, for example, is taught this way.) Hiring a few AI experts to teach in person can also motivate employees to learn these AI skills. AI will revolutionize many professions. You should tell everyone that they need to find their niche in the AGE of AI. Consulting with experts can help you tailor a course to suit your team. However, a nominally educational plan might look something like this:

  • Supervisor and senior business manager :(training time: 4 hours)

Goal: To get the executive to understand what AI can help the company do, to start developing AN AI strategy, to make appropriate allocation decisions, and to work with the AI team that supports valuable AI projects. Courses:

  • Understand basic AI concepts, including basic technologies, data, and what AI can/cannot do.

  • Understand the impact of AI on company strategy.

  • Case studies of AI applied to similar industries or your industry.

  • Department leaders who perform AI projects :(training time: 12 hours)

Goal: Department heads should be able to set direction for AI projects, allocate resources, monitor and track progress, and make adjustments as needed to ensure successful project delivery. Courses:

  • Understand basic AI concepts, including basic technologies, data, and what AI can/cannot do.

  • Understand basic AI techniques, including the main categories of algorithms and their requirements.

  • Understand basic AI project workflow, AI team roles and responsibilities, and team management.

  • AI engineer trainee :(training time: 100 hours)

Goal: Newly trained AI engineers should be able to collect data, train AI models, and deliver specific AI projects. Courses:

  • Deep understanding of machine learning and deep learning techniques; Basic understanding of other AI tools.

  • Learn about the available (open source and third-party) tools for building AI and data systems.

  • Ability to implement AI team workflow.

  • Also: keep learning to keep up with AI technology

4. Build an AI strategy

An AI strategy can lead your company to create more value, but it can also build defense mechanisms. Once your team has seen the success of your initial AI project and deepened your understanding of AI, you can identify where AI can add value and focus on that.

Some executives will argue that building an AI strategy should be the first step. But in my experience, most companies don’t have a well-thought-out AI strategy until they have some basic experience. These basic lessons can be learned from the previous three steps.

As AI evolves, your approach to building defensible moats will change. Here are some ways to consider:

  • Build multiple different AI assets based on a unified strategy: AI enables companies to build unique competitive advantages in a new way. Business Strategy by Michael Porter shows that one way to build a barrier business is to build multiple different assets according to a unified strategy. And make it harder for competitors to replicate them.

  • Use AI to build industry-specific advantages for your company: Rather than competing with a tech company like Google on “broad” AI, I suggest you become the leading AI company in your industry. Developing unique AI capabilities will give you a competitive advantage. The impact of AI on your company’s strategy depends on the industry and context.

  • Design positive feedback strategies based on the “virtuous circle of AI” : In many industries, we can see that data accumulation creates barriers to business:

For example, web search engines such as Google, Baidu, Bing and Yandex have vast data assets related to user clicks and search terms. This data helps these companies build more accurate search engine products (A), which in turn helps them get more users (B), which in turn helps them get more user data (C). This positive feedback loop makes it hard for competitors to break through.

Data is a critical asset for an AI system. As a result, many AI companies also have sophisticated data strategies. The key elements of your data strategy should include:

  • Strategic data acquisition: Useful AI systems can be built from 100 data points (small data) to a billion data points (big data). But more data can only be more helpful. AI teams use complex, multi-year strategies to capture data, and strategies to capture data vary from industry to industry and context. Google and Baidu, for example, have plenty of free products that give them access to commercially valuable data.

  • Unified database: If your database is controlled by 50 different executives or departments, it is almost impossible for engineers or AI software to access the data and connect to nodes. Instead, centralize the data or aggregate it into a small number of databases.

  • Learn to differentiate between high and low value data: just because you have terabytes of data doesn’t mean the AI team can create value from it. Expecting an AI team to magically create value from a big data set is likely to fail. I have mourned the sight of ceos overpaying for low-value data, or even buying a company for it, only to discover that the target company’s megabytes of data were useless. To avoid this mistake, start building an AI team at the beginning of data collection and let them help you decide the priority of the data to be collected and stored.

Create network effects and platform advantages: Finally, AI can be used to build more traditional “moats.” For example, platforms with network effects are highly defensible businesses. Their inherent winner-take-all nature forces companies to grow fast or die. If AI allows you to acquire users faster than your rivals, then you can use AI to build a “moat” that can be defended by the platform’s aforementioned features. More broadly, you can also use AI as a key component of a low-cost strategy, a high-value strategy, or other business strategy.

5. Establish internal and external communication

Ai will significantly impact business. If it has affected key stakeholders, you should ensure that progress is consistent by running relevant communications procedures. Here’s what you should consider for each audience:

Investor Relations: Leading AI companies (Google, Baidu, etc.) are also more valuable today, in part because of their AI capabilities and the impact of AI on their performance. By creating a clearly explained value creation thesis for the AI of your business, describing your company’s growing AI capabilities, and finally presenting a mature AI strategy, investors will be able to properly evaluate your business.

Government relations: Companies in heavily regulated industries (self-driving cars, health care) face unique challenges in maintaining business compliance. For such companies, building a credible and compelling AI vision, and explaining how your project can bring value and benefits to the industry or society, is an important step in building trust and goodwill with governments. These suggestions need to be combined with direct communication with the government and an ongoing dialogue with regulators as you roll out your corporate projects.

Customer/User training: AI can be a huge benefit to your customers, so be sure to communicate proper marketing and product roadmap messages.

Talent/Recruitment: With ai related talent scarce, a strong employer brand will have a significant impact on your ability to attract and retain such talent. Ai engineers often want to take on exciting and meaningful projects. So, as an employer, showing success in your business will help you recruit.

Internal communication: At present, the public still knows little about AI, and there is fear, uncertainty and suspicion in the public mind due to the over-hyping of strong AI. Many employees also worry about losing their jobs to AI. Although this perception varies from culture to culture (the sense of panic is worse in America than in Japan, for example). Therefore, clear internal communication can not only explain AI thoroughly, but also reassure employees and reduce internal resistance to the adoption of AI.

It is vital to your success to follow the rules of history

Understanding the transition in the era of the rise of the Internet makes sense to guide companies toward AI. A lot of companies made a mistake during the rise of the Internet, and hopefully you can avoid that mistake during the rise of ARTIFICIAL intelligence.

Here’s what we’ve learned from the Internet age:

Shopping center + Website ≠ Internet company

Even if a shopping mall sets up a website and sells goods on it, that does not, by itself, turn the mall into a true Internet company. The real definition of an Internet company is: can you use the Internet to its advantage in your company?

For example, the A/B test commonly used by Internet companies, which periodically launches two versions of A website and compares which works better. Tech companies can even run hundreds of experiments simultaneously, but that’s hard to do in a physical shopping mall. An Internet company can launch a new product every week and learn from competitors very quickly, whereas a shopping mall may update its design once a quarter. There are unique positions such as product manager and software engineer in Internet companies, and these employees have a unique form of collaboration and workflow.

Deep learning is one of the fastest growing directions in AI right now, and it has some similarities to the rise of the Internet. Today we will find out: any ordinary company + deep learning technology ≠AI company

In order for your company to be good enough at AI, you need to guide your company to take advantage of AI’s true strengths.

In order for your company to fully transition to AI, you must:

  • Systematically execute multiple valuable AI projects: AN AI company must have outsourced or proprietary technology and talent that can systematically execute multiple AI projects and directly impact the business.

  • Good UNDERSTANDING of AI: Company employees need to have a general understanding of AI and have processes in place to systematically identify and select valuable AI projects.

  • Steer the strategic direction: The company’s strategic needs are broadly aligned with the ai-powered future.

Turning a large company into a strong AI company is challenging, but it can be done with the right partners. Landing.AI is dedicated to helping partners transform their AI businesses, and the company will share more practices in the future.

Ng estimates that it typically takes two to three years for traditional companies to transform their AI businesses, but people can see the initial results after six to 12 months of implementation. Companies investing in AI will grow faster than their competitors.