The rapid advancements in Artificial Intelligence (AI) technology are not only prevalent across the entire industry chain, encompassing everything from chips, terminals, algorithms, system platforms, to application services, but are also proliferating throughout various industries and domains. This proliferation is reshaping the challenges and opportunities for AI application development, and even becoming one of the indicators for measuring the competitiveness of countries, companies and individuals.
Given the rapidly changing AI landscape, proper use and governance of AI have become a critical issue for many companies and organizations. IEK Consulting predicts that over the next two to three years, AI will develop on four fronts: distributed AI, generative AI, trustworthy AI and sustainable AI, which is the ultimate goal.
In the midst of the swiftly evolving AI wave, how to harness and govern AI effectively has become a crucial consideration for many businesses and organizations. IEK Consulting predicts that over the next two to three years, AI will evolve in four major directions: " Distributed AI," "Generative AI," "Trustworthy AI," all aiming towards the development goal of "Sustainable AI."
Distributed AI: Edge AI 2.0 Emerges as the Key to Edge Collaboration
In recent years, the evolution of Edge AI has extended beyond the realm of AI chips, encompassing AI algorithms for the processing and analysis of data generated or collected on devices and across the internet. This shift aims to address the imperative of reducing the training time and resource demands, including computational power, time, and data, required by AI models. As we navigate these transformations, the role of Distributed AI becomes paramount, serving as the linchpin for orchestrating collaborative efforts at all edges.
Distributed AI aims to establish edge AI that is more real-time, more reliable and with better privacy and security protection in order to achieve faster data collection, analytics, and responses. Edge AI 2.0 emphasizes the deployment of AI algorithms on the edge (or devices) to shorten AI model training time, accelerate the speed of AI model inferences, reduce the computational power consumption of AI and boost the effectiveness of edge orchestration.
Generative AI: Rapidly Evolving Use Cases and Reshaping All Industries
The emergence of Generative AI (GAI) become the hottest topic recently. Generative AI can create a variety of new content based on existing data or patterns. Support by Large Language Models (LLMs), GAI can engage in cross-modal training and learning on massive datasets, generate novel types of content such as texts, images and voices. As Generative AI tools gain widespread attention, they are reshaping industries across the board.
GAI applications can be categorized into four main aspects: (1) automated data generation and synthesis. The proportion of synthetic data will grow rapidly as the result of global GAI development. This will also resolve the current problem of data scarcity. (2) The wide range of GAI functions will enable accelerated product or service innovation, especially in the development of new drugs (e.g., protein folding) and the development of new environmentally-friendly materials. Manufacturers will also be able to improve the efficiency of product development and speed up the establishment of virtual factories. (3) GAI lowers the threshold of program design and development and will result in the mushrooming of GAI tools for applications at work and in life. It will replace humans for repetitive tasks and boost productivity with man-machine cooperation. It will also change the ways of working, enhance the value of human input, and allow people more time to pursue a better quality of life.
Trustworthy AI: AI Credibility Issues Will Drive the Emergence of AI Assessment Tools
The advancement of GAI has also brought about the "Deepfake" phenomenon. For instance, using AI, it's now possible to replicate almost 50% of human faces worldwide, making it challenging to distinguish between authenticity and deception. This proliferation of misleading content has led to various issues concerning individuals, businesses, societies, and countries. Consequently, the development of Trustworthy AI has become a significant focus.
Industry players, governments, academia and research institutions around the world have begun actively embracing Trustworthy AI. The European Union, the OECD, the U.S. and Singapore are the cases in point. International corporations like IBM,which have developed Trustworthy AI toolkits and entrusted them to the Linux Foundation for oversight.
A series of workflows from data collection, data cleaning and analysis to machine learning and deep learning have been put onto the platform. At this juncture and amid the frenzy for GAI, LLMs, ChatGPT and others, concerns about the accuracy of AI-generated data, algorithmic biases, model misuse, security, and privacy protection have highlighted that Trustworthy AI has become an important issue for individuals, companies, societies, and governments, requiring serious attention or involvement.
The Industrial Technology Research Institute (ITRI) focuses on the research and development of Trustworthy AI technologies, starting from the assessment and testing of systems and services. This approach aims to ensure AI data quality and compliance, model and system security, performance and robustness, fairness and explainablility, and more. This effort assists in enabling AI technologies, products, or services in Taiwan's industry to meet international standards. It supports domestic businesses in exporting various types of AI application solutions overseas, such as generative AI-related applications, facial recognition, fingerprint recognition, voice recognition, automotive applications, medical imaging, and other AI products, systems, or services.
This enables domestic enterprises to become trusted partners in the international industrial chain, facilitating swift entry into global markets and the exploration of new AI business opportunities.