According to Sami Viitamaki, Havas Digital Executive Director of advertising company VentureBeat, the company will need to hire the chief machine intelligence officer (CMIO) in 2017. The article details relevant challenges such as business needs, customer value, and autonomy and interactivity. The following is the main content of the article: Machine Intelligence (MI) is quickly out of the cradle of the technical department. This requires companies to have a new executive function and strategic roadmap. In this article, I will outline some of the areas that your new chief machine intelligence officer (“CMIOâ€) will overcome in 2017 and beyond. Currently, the machine is learning how to run a physical retail store, starting to outwit medical doctors and even inventing their own private language. Therefore, it is natural for machine intelligence to ask questions and issues. I have been fortunate to have engaged in and supervised a lot of machine intelligence projects, and I have also been deepening research in this area. I think that as long as the interests and challenges are taken seriously, the technology will outweigh the disadvantages for both companies and individuals in the future. Artificial intelligence will take over as the chief machine intelligence officer in the service industry. One of the important indicators for measuring social development is its productivity, and machine intelligence will become a huge productivity driver. For example, doctors, lawyers, or their back office team can take weeks or even months to complete. Watson can be done in seconds or minutes; Amazon’s Amazon Go stores use machine intelligence. Provides an ultra-fast retail experience with skip-the-line checkout; Google’s machine intelligence has been able to translate languages ​​it has not encountered before; Salesforce’s enterprise machine intelligence can significantly simplify administration, reporting and coordination efforts, knowing that these jobs now occupy more than management 50% of the time. All this will give people more time to think, manage other people, and do what they enjoy. Undoubtedly, as with all other technological changes, machine intelligence will also bring challenges, and it may also make the Internet revolution seem insignificant. However, although it is widely considered to be the key to future success, companies often have not explored details and a comprehensive road map. I think it should be the strategic focus of any big company. It should be led by an executive who reports directly to the CEO. These CMIOs will assume the task of bringing all business units and the company as a whole to the age of machine intelligence. If you are one of them or want to recruit a CMIO, here are some of the major CMIO-level challenges you will face. Business Needs Where are the hidden links to be discovered? What important aspects should you implement to diagnose and predict? What volume of daily tasks should you automate? Where should you use rapid, scenario-based customization and real-time optimization on a large scale? These are some of the most promising areas of the value of machine intelligence enterprise application. The CMIO will need to assess the greatest opportunities and the most urgent issues by fully understanding the company's environment, business processes, and the relationships between them. In some areas, machine intelligence won't even bring a competitive advantage - it's the price that business must pay. Throughout history, the speed of adoption of new technologies has accelerated dramatically. Today, our mobile Internet, easy-to-access cloud services, and the Internet of Things are all in place. In this context, machine intelligence can take over all walks of life more quickly than previous technologies, leading to a profound change in the operating model of the company. For example, when we no longer browse search results, just ask questions to get a correct answer and fast implementation, how will Google make money? The laggards will face difficulties or even be completely eliminated. The traditional enterprises that are standing still will encounter similar situations under the impact of rivals born on the Internet. However, the impact of this machine intelligence technology will be even more significant. Customer Value In addition to machine intelligence, design thinking and customer experience management will become important concepts for business success in the near future. People are not rational calculators. They often value soft experience factors more than hard objective values. Machine intelligence has great potential for providing people with more valuable experiences. CMIO needs to explore these opportunities through cooperation with design departments and agencies. Machine intelligence simplifies everyday tasks, makes it impossible to do what was previously impossible, teaches people, personalizes products and services, promotes important connections between adults and objects, and makes experiences safer. Bring new storytelling and entertainment. These are just some of the areas of the innovation experience, and in all these areas, the seamless transition between platforms and devices will be the key. Machine intelligence will involve more than just "what". The upcoming book by my friend Alexander Manu will focus on this insight: In the behavioral economy, delivery really is the entire value proposition. Chat bots and cognitive personal assistants are already making many common tasks smoother and more automated. They can understand their owners anytime, anywhere. Also note that you have both external and internal customers. Facing internal applications will usually be feasible soon. The problems in this area are relatively simple and there are enough data available. In areas where the problem is more complex and fuzzy, where data is yet to be explored or lacks, companies often should establish decision-making and support systems for employees. Their experience with customer and customer data will be greatly improved. Autonomy and Interactivity The question of how much autonomy is given to machine intelligence solutions in dealing with customers, employees, and other stakeholders is a key issue. Ideally, intelligent systems should handle autonomous tasks completely autonomously, but for some tasks, even if they are competent, people may not be able to trust them. Google’s several AI products, Bob, Alice, and Eve, found that they communicated with each other in a language that humans could not understand. This made some people very worried. In addition, robots that have had crazy trading in the past have caused the market to crash, causing investors to suffer huge losses. Nick Bostrom classifies machine intelligence into three levels in his book Superintelligence: Oracles tells users what they need to know, just as Google search has done for decades. Like that; Genies can satisfy the owner's wishes, but only according to their requirements, just like the current Siri; Sovereigns only needs a few important principles to be able to operate independently on behalf of the owner. Although true Sovereigns do not exist yet, the portfolio of robots created by brokers today can be said to be relatively close. Regardless of the degree of autonomy provided to machine intelligence, the CMIO still needs to lead development efforts and monitor the performance of the company's machine-intelligence interaction practices. Machine intelligence systems increasingly interact with users through text and voice. Market research firm Gartner predicts that within 4 years, there will be 30% of searches that do not require a screen to complete; the first is through speech, and the latter may pass through the brainwaves. Although dialogue is the most common form of human interaction, it is not necessarily the most natural interactive medium for brand owners. In addition to logos, promotional slogans, and other common brand assets, the essence of the brand will be extracted from questions, answers, small interactions, emotions, and conversation styles. These will become important business and brand factors, measurement and analysis technology will need to evolve to reflect these subtle differences. Technical Architecture An important task for CMIO and CTO (Chief Technology Officer), COO (Chief Operating Officer) and CFO (CFO) is to come up with the technical architecture of machine intelligence and the roadmap and investment plan. The machine intelligence ecosystem is already a full-blown, fast-growing garden: with both cloud services and on-site solutions, and hardware and sensors for large and small businesses that are tailored to the core functionality and support functions and customized To cater to all walks of life. The technology stack also includes several vertical levels ranging from understanding natural language and complex concepts to data collection, data science and machine learning, to multi-purpose libraries and machine intelligence training venues. There are already many companies offering out-of-the-box features and APIs, and there are plenty of free or cheap resources for you to build your own solutions from scratch. Although open source free libraries like Google's TensorFlow may sound great, developing internal machine intelligence at the beginning requires thinking twice. Although those basic tools may be free, you still need an experienced team of scientists and experts to develop smart models and algorithms that you can improve on your own. This is not easy: The relevant talents are not easy to find, let alone the high costs of hiring them. The StackOverflow survey found that only 0.1% of respondents think they are machine intelligence developers and only 1.9% have research background in mathematics or statistics (important skills in machine intelligence development). Even with easy-to-use interfaces and options to train machine intelligence, you still need experts to fix it when you encounter problems. Therefore, rather than trying to defeat Google, Salesforce, Amazon, Microsoft, IBM, and Facebook in a global talent war, it is better to establish cooperation with these companies. Moreover, integrating the giant's machine-intelligence stack with the tools you already use will bring additional benefits. Especially for subdivided specialized use, small but perhaps very suitable provider choices are becoming more and more. But also keep in mind that in the field of machine intelligence, the scale is usually a good one, after all, bigger data means faster and better learning opportunities. System Learning The machine intelligence system needs continuous learning, because at the beginning of the launch it was just a “babyâ€. In fact, researchers are creating virtual babies to help understand how babies learn and teach computers to learn faster. It is an important task of the CMIO to develop learning objectives and strategies for the enterprise machine intelligence project. Continuous data flow plays a crucial role in the learning of machine intelligence systems. The rule of thumb for machine intelligence data is “more is betterâ€. The main advantage of machine intelligence is that it does not require a sample - it can learn directly from all available data, thus making insight and prediction more reliable. The theory of how data and questions may be related is good, but keep in mind that as long as there is enough data available, machine intelligence can help establish new connections, discover new opportunities, and answer questions that you haven't even thought of. Different issues will require different learning paradigms. Reinforcement, supervised learning, and unsupervised learning are the most widely known machine learning methods of the moment. There is also already an ecosystem for training purposes: Open AI of Elon Musk has just opened the Universe, Universe provides tools for everyone to train, measure and evaluate intelligent systems; OpenAI Gym is used for training reinforcement Algorithms; Microsoft has open-sourced the Microsoft Cognitive Toolkit to accelerate the development of machine-intelligence applications and is making progress. Even Apple’s most mysterious technology company has also opened its machine-intelligence research to the academic community. It is clear that the future of machine intelligence will depend on collaboration and collective learning. A good idea to think about smart system training is to gamify it and introduce triggers, actions, failures, successes, and rewards and feedback loop elements that make it easy for people and machines to understand. Research firm L2 appropriately sums up the importance of learning with a formula that represents the future value of smart systems: the number of available data sensors × actionable intelligence that can be fed back to the system = value proposition. The final part of the artificial part is that the CMIO will need to address the artificial part of the smart system because machine intelligence cannot yet operate independently. People only need to provide machine intelligence systems with data clues that they can measure, learn and apply to new scenarios. However, they also need to manually check data, adjust algorithms, provide training sets, provide guidance for machine intelligence in correct and incorrect interpretation, and manage situations where machine intelligence cannot resolve independently. Reinforced learning and supervised learning require more human involvement, while unsupervised learning allows the machine to learn on its own. However, unsupervised learning is the most experimental of these three, and such systems cannot yet be created by themselves. The Google AI Experiments trial is a good example of how everyone on the Internet can be used to play smart systems and train them in an open environment. One thing to keep in mind is that machine intelligence systems are still built by people, so if they are not properly controlled, they will retain their creator's prejudice. This involves not only the scientists who built the algorithm, but also the annotators who annotated the training set. For example, if the highest score is 5 stars, the lovely index score that women give to the same puppy is on average 0.16 stars higher than that of men. This is a statistically significant difference. It is of great importance in the teaching of machine intelligence, such as the concept of teaching cuteness. Even with a near-perfect system, the CMIO still needs to decide to what extent it trusts those machines, how much control it retains, and how it establishes checks and balances. As mentioned earlier, one of the great promises of machine intelligence is that it will allow your staff to focus more on humanistic work such as judgment, social contact, and creativity. Over time, the need for control should be reduced. Summary At Havas, we currently carry out many marketing and product marketing projects that use machine intelligence around the world and in all walks of life. Perhaps the best thing that impressed me was the "Brothers Who Trusted Home Team" program that we launched with our customers TD Ameritrade and IBM Watson. It brought a new perspective on TD Ameritrade's NFL sponsorship. For the football season, we created an attractive web application to measure the relevance of football fans’ confidence in their home team on the social channel and the home team’s success. The industry's first solution is a surprise. For example, it accurately predicted the results of all the NFL Champions team Denver Broncos, including the generally unpopular Super Bowl game of the Broncos. Machine intelligence will take over many industries and vertical areas in the coming years. The question now is not whether this will happen or when it will happen, but rather how much preparation do you have for the arrival of the age of machine intelligence? Considering machine intelligence as an important strategic division within the company is a good start.