The Secrets of AI, from Theory to PracticeDec 21, 2018
Artificial Intelligence (AI) is one of the most ubiquitous buzzwords these days. From theoretical AI to Enterprise Intelligence (EI), one question remains: how do individuals and enterprises use AI to unlock the secrets of growth and development to turn academic research into practical use?
How does AI influence people's lives?
David watched the hit TV show Westworld before going to bed last night, imagining a world highly influenced by AI. In the morning news, he read reports that AlphaGo, a Go-playing computer program, had bested a number of Go masters. At this moment it suddenly dawned on him how much AI is used in the real world.
Going to his job as a logistics manager today, David learned that the overall layout of the warehouse had changed overnight; and over the course of the day he noticed it was far more convenient to pick up goods than before. AI had entered David's workplace and immediately become a strong ally for him by cutting his workload by almost half.
AI is not so far away from any of our lives. The Enterprise Intelligence (EI) service on the HUAWEI CLOUD helps enterprises optimize their warehouse systems. After one instance of optimization, the walking distance of picking clerks was reduced from 30,000 steps to 20,000 steps, and the pickup efficiency increased by about 30 percent.
The bottom line is that AI is influencing our daily lives.
Many experts have divided the implementation of AI technology into three phases:
In the first phase, AI technology becomes mature and starts to achieve goals that exceed the average human level in multiple fields.
In the second phase, enterprises will use AI as a necessary resource like fuel and electricity.
In the third phase, general AI will directly affect people's everyday lives, somewhat resembling a society we have imagined in movies and television shows.
At present, AI is in the critical period of transit between the first and second phases, and, as such, has many enterprises scratching their heads, wondering how to apply AI technology to meet their requirements.
According to the Global Trends Study from TCS, a world-renowned consulting organization, 84 percent of the world's surveyed companies regard AI as a key factor in core competitiveness.
What are the challenges for the development of AI to EI?
Review: Development of AI
In 1945, British mathematician Alan Turing proposed the concept of an Automatic Computing Engine. In 1956, the first AI program was created during the Dartmouth Summer Research Project on Artificial Intelligence.
According to the definition provided by the Alan Turing Institute, AI technology in use today is a series of technologies that imitate human thinking, perception, and communication. In other words, any computing and statistical technology that mimics human intelligence can be called AI.
Through the 1960s, early AI applications supported by Turing's theories on computational logic mushroomed, though proved difficult to reduce to any practical use and by 1974, the first AI development bubble had burst. The prospects for any form of industrial AI remained largely stagnant for several decades to follow.
The resurgence of AI today is the product of a revolution in computer science that has focused on machine learning technologies that implement Natural Language Processing (NLP), machine vision, knowledge mapping, deep learning, and reinforcement learning. With the dividends brought by Internet technologies and Big Data, including cloud computing and new chip technologies, in recent years AI has developed quite rapidly.
Trio: Gifts from AI to Enterprises
According to the 2017 article Growing the Artificial Intelligence Industry in the UK, released by the UK government, AI will contribute up to USD 15.7 trillion to the global economy in 2030, which is more than the current economic output of China and India combined. Of this figure, approximately USD 6.6 trillion may come from increased productivity and USD 9.1 trillion from consumption effects.
How is it that AI became so promising?
In general, AI technology driven by machine learning and neural networks meets the growth needs of enterprises at three levels.
1、Development of interaction mode
AI technologies, such as NLP, voice interaction, machine reading comprehension, machine vision, and machine sensing are changing the relationship between humans and machines. Finger commands are clearly not the only way for people to interact with machines. It is now common for humans to use voice or hand and body gestures to deliver input commands to machines.
In one example, the training and attitudes of shop assistants who communicate directly with customers in brick-and-mortar retail chains will affect sales revenue. By analyzing customer shopping habits through in-store cameras using machine vision technologies such as facial recognition, retail enterprises are learning to infer the satisfaction level of each customer.
Another reason AI is influencing the strategic direction for many enterprises is that AI can be used to mine enterprise data for otherwise hidden trends. Through the optimization of data structures and the use of algorithms to intelligently calculate more efficient production models, enterprise operations are moving from manual, experience-based judgments to AI-based deductions based on real-world data.
In the area of asset inspections, pre-AI enterprises would conduct manual reviews and draw conclusions based on human assessments. Compared with modern methods, the inspection efficiency was quite low and the results were necessarily reported with significant margins for error. There are no such problems in the EI era, as image recognition technologies help to accurately and efficiently count and track inventory.
Another simple example is a cattle ranch. In the past, if you needed to count the herd, you needed to count the animals one-by-one. The introduction of AI improves efficiency tremendously by allowing the number of animals to be counted by machine after photographing or video recording of the farm from a camera-equipped drone.
AI technology helps enterprises improve yield rates and production-safety coefficients while reducing labor costs. In addition, enterprises are using AI-based tools to improve the work environments for employees and while reducing overall costs.
Using the example of the internal audits that commonly occur in enterprises, traditional manual collection methods are time-consuming and error-prone. An alternative is the use of the Optical Character Recognition (OCR) function hosted on the HUAWEI CLOUD platform to automatically digitize, recognize, and catalog employee invoices to save labor costs and improve efficiency.
In addition, AI applications help reduce social costs, increase social efficiency, and promote environmental protection. For example, Huawei has provided a smart heating solution for a heating company that uses a reinforcement-learning technology. Using the AI, the generation of heat is adjusted precisely according to the direction and rate of change in temperatures both indoors and outdoors. The result has been a 10 percent reduction in fuel consumption while maintaining an indoor temperature of 70°F (21ºC). In the face of today's severe environmental challenges, intelligent technologies are increasingly crucial to stimulating future energy conservation and environmental protection plans.
However, the road to intelligence will not be built overnight.
Challenges: A Long Way from Theoretical Research to Commercial Use
As the AI era gets ever closer, many entrepreneurs, enterprise managers, and technical teams want to use AI to complete their enterprise transformation, but they may encounter various problems in actually doing this.
In the annual report Reshaping Business with Artificial Intelligence issued by the Massachusetts Institute of Technology (MIT), analysts point out that many enterprises wish to carry out enterprise transformation using AI, though only about one in five has actually incorporated AI into their products or processes.
From introduction to application, enterprises may face the following difficulties in AI-based transformation. Chinese enterprises report unique challenges during this period of opportunity:
1、Imbalance between supply and demand
Current practice by large numbers of Chinese enterprises is to access AI using only basic machine learning models. The technical solutions acquired from some technology vendors appear sophisticated but will not meet the actual needs of their customers. Simple machine learning models may prove to be a waste of enterprise money and hard work where enterprises will reap half the rewards for twice the effort.
According to Profiles in Innovation: Artificial Intelligence - AI, Machine Learning and Data Fuel the Future of Productivity, released by Goldman Sachs, China accounted for 51 percent of the global emerging AI projects in 2017, surpassing the United States. For AI talent, however, China only makes up 5 percent across the world. The huge talent gap and the demand for talent make it difficult for Chinese enterprises to find individuals that truly understand AI and are able to build and execute EI projects.
If an enterprise wants to implement AI-based transformation today, or try to integrate AI into its services portfolio, it must bear the costs that arise from executing the algorithms at scale, including computing resources, data purchases, talent and team expenditures, and equipment expenditures. In the absence of a clear revenue program, it is little wonder that many enterprises are hesitant to invest in AI.
Pressure from many sides, as well as uncertainty about the future, have significantly slowed the pace of AI entering enterprises; however, as technical upgrades and platform-level enterprises continuously improve services, more Chinese enterprises will benefit from AI.
Opportunities: Next Stage of EI
Today, enterprises use AI to complete their own business transformation, achieve industry competitiveness, and seize opportunities from different aspects.
EI has the following advantages:
1、Continuous improvement of enterprise-class services
In countries such as the UK and Israel, enterprise-class service projects occupy more than 85 percent of total AI projects, while the ratio in China is quite low. Fortunately, as China's cloud computing, Internet, and other high-tech enterprises enter the AI market, algorithm and data providers, and enterprise-class service providers also share the market. Today, enterprise-class AI services are booming in China's market. Cloud service providers develop various AI services for enterprises. For example, the HUAWEI CLOUD has released more than 30 EI services.
2、Deep integration of AI into various industry applications
AI can deal with problems of various traditional, small data, and specialized industries.
For instance, in recent years 'intelligent agriculture' has become a hot topic. AI is a new force that escalates the performance of agriculture, animal husbandry, and even environmental governance. As with agriculture, AI is on the verge of penetrating other major industries and professions, such as engineering, retail, government, transportation, and public service.
AI services and products are now affordable as never before. Data and computing resources required for machine learning are also within the reach of enterprises in various fields in different ways.
Cloud computing service providers are promoting AI transformation. In the past, enterprises that wanted to use AI had to understand the technology enough to employ a team of AI specialists. But now the situation has changed. Google, for example, has released AutoML, an automatic machine learning system that allows enterprises to gain access to AI capabilities in some areas without compiling code. The Huawei Cloud is also promoting high-quality AI services that open AI to enterprise customers that operate on tight budgets for external services. Huawei Cloud eliminates the barriers to entry to allow enterprises to quickly engage in the intelligent era.
AI is being popularized as an enterprise necessity, just like electricity, computers, and networks had been in years past.
From Intelligence to Practice: The Secrets of AI in Enterprises
It should be noted that in the process of AI becoming practical for day-to-day operations, that many enterprises remain unclear or unpracticed about how best to apply the technology. In this early period, there are many enterprises that have invested in AI but so far show little benefit for the time and money spent.
In the face of new trends, real difficulties, and future values, there are better or more disciplined ways for enterprises to enter the AI world. There are also related precautions. These could be called 'the secrets of AI.'
1、Quickly accessing AI through cloud services
The technical features of machine learning point the way for using cloud-computing platforms as the easiest way for enterprises to access AI for their business operations. Each of the major cloud computing platforms is competing for enterprise market share by offering the most appropriate, cost-effective, compatible, and customizable AI channels. The adoption of one or more of the available cloud-computing ecosystems is the main way for enterprises to invest in the AI world today.
Obtaining AI products and services through cloud platforms like HUAWEI CLOUD is efficient because enterprises can conduct AI-based transformation very efficiently and at the lowest possible cost. Because AI is highly modularized it is not at all necessary for any enterprise today to start from scratch.
2、Adopting mature EI solutions
For most enterprises, the best way to avoid a mismatch between technology and reality is to look for verified AI solutions that fit their own needs. Because many basic enterprise requirements are similar, there are multiple enterprise success stories that can be used as reference cases for minimizing risk.
For example, the EI services provided by the HUAWEI CLOUD incorporate the many years of Huawei's development achievements in the carrier and enterprise markets. The EI solutions provided by the HUAWEI CLOUD are verified from the company's long history of meeting or exceeding customer requirements. This level of trust is crucial to Chinese companies that are considering the risk-reward proposition for any new EI solution.
3、Escalating EI by leveraging academic outputs
AI relies heavily on the combination of academic research and practical use. Continued, broad information acquisition is the only way to ensure that enterprises are at the forefront of technology and given the opportunity to embrace efficient solutions.
Today we can see new academic achievements being integrated into many industries very quickly. The feedback loop to academia becomes a direct driving force for further industrial development for enterprises. The enterprises themselves are setting up academic research teams to fill the basic need for effective, targeted AI and EI solutions. The main challenge remains of finding the most efficient ways to integrate academic research into practical scenarios.
Challenges and opportunities
Many competitors are entering the AI market to offer solutions to enterprise customers.
For all practical purposes, Chinese enterprises are faced with a wide range of business prospects, technical advantages, abundant platform choices, and a good strategic environment nationally. At the same time, real limits exist, including a shortage of skilled talent and lack of understanding where and how to apply the tools. In other words, a steep learning curve will restrict the rate at which enterprises will be able to develop their own intelligent services. AI is powerful tool with the potential to promote development and unlock great opportunities. Enterprises should learn to be masters of AI rather than slaves to it.
For the AI technology vendors, the key factors for enterprise growth remain rooted in the fundamentals of access to academic research, investment capital, and aggressive research and development. The goal is to maintain a high degree of industry cooperation and allows enterprise customer to experiment quickly and inexpensively in their effort to absorb new trends and technologies.
Only in this way will AI truly drive global industrial development and help enterprises achieve long-term growth in a fast, environment-friendly, and cost-effective way.
Compared to the historical development of AI, the new world of EI is bringing practical values to enterprises by eliminating technical barriers. Putting AI into everyday use will become one of the recurring main trends for the evolution of future AI ecosystems.