According to the global market forecasts for artificial intelligence (AI), the revenue of the global AI industry will reach US$1.57 trillion by 2030, on an accelerating upward trajectory. The development of the AI industry is gradually converging around automated services based on machine learning, edge AI, and trustworthy AI.
The first trend is for automated services based on machine learning, with a focus on lowering technical barriers, learning with limited data, and rapid modeling. The pursuit of high AI penetration has seen a mushrooming in the number of market players. The global market for AI software, such as machine learning, natural language processing, and Robotic Process Automation (RPA) is expected to reach US$22.6 billion in 2020. Heavyweights such as Google, Amazon, and Microsoft and start-ups including H2O.ai and Data Robot are all entering the market for “Democratizing AI”. The involvement of both large companies and start-up firms is accelerating the demand and the development of edge computing.
The growing demand for low-latency applications over recent years, the overly high price of cloud transmission, and the privacy concerns of users have contributed to the development of edge computing. It has been widely adopted by companies selling platforms, systems, and chips. Edge computing has been used by component and IoT equipment vendors, system integrators, and start-ups. "Serverless computing" and "machine learning computing" that are highly flexible, scalable and real-time responsive are the hottest areas in edge computing.
"Trustworthy AI" is another key development trend for the AI industry. The AI black-box approach to operation and problem solving has been the Achilles heel in the development of AI. Major decisions made by AI are too often not adopted by humans due to concerns about incomplete modeling and data. The determination of the road status by autonomous vehicles and the medical decisions made by AI regarding surgery and treatment all raise doubts in the mind of the user about the accuracy of analysis by AI. This is the reason why the U.S. Defense Advanced Research Projects Agency (DARPA) established the Glass Box model, in order to create an explainable AI architecture where input data and output results can be monitored. The trustworthiness of AI is established through monitoring by humans.
PwC conducted a survey of 1,000 companies in the U.S. in order to analyze the AI demand from different industries. The top three priorities reported were “enhancement of AI safety via validation, monitoring, and authentication”; the construction of transparent, explainable, and provable AI models; and the establishment of ethical, understandable, and legal AI systems. This demonstrates the foremost corporate demand is for trustworthy AI.
In sum, AI for industries should address their specific pain points and meet the requirements of different industries by developing customized solutions. Currently, traditional industries, retailers, manufacturers, service operators and ecommerce platforms all have their own specific AI requirements.
On the other hand, industries for AI should seek to resolve the universality issue in order to scale up the market. This may be viewed as a new type of AI technology services. It is applicable to the development of technology platforms, R&D testing, technology consultancy, system integration, or software development. In essence, it is AIaaS (artificial intelligence as a service).
How can Taiwan be at the forefront of this new wave? Early preparations are to be advised for a diversity of innovative concepts by integrating AI hardware and software.
Taiwan may develop AI in automation by combining AI with IoT (Internet-of-Things) in smart manufacturing, autonomous driving, or drones to achieve full automation. This is also what Taiwanese industries are best at. Another focus should be "distributed AI", to ensure real-time responsiveness, reliability, stability and safety of computing and analytics. This enables the completion of all tasks on the edge to meet corporate needs with better efficiency and risk diversification. This is then followed with "trustworthy AI" and data optimization so that algorithms, systems, and business models are all explainable, traceable and safe.
Finally, AI may assist the transformation of companies in Taiwan by enhancing industry resilience, with distributed and ultra-automated products/services offering security and privacy for the AI market. This may be the provision of quality, insightful, and shareable "data services", or low-cost, highly universal, and automatic sensing "edge terminals". Alternatively, it may be possible to provide ultra-automated, explainable, and edge learning "algorithm services", or develop the "computing power" with low energy consumption, high efficacy and scalability. These are all strengths that Taiwan can deploy in the development of its AI industry.