A wave of artificial intelligence (AI) has swept across the globe. The World Artificial Intelligence Conference 2018, held in Shanghai this September, drew thousands of participants from 40 countries and regions, as well as hundreds of big and small enterprises. Applications of AI to nearly all sectors of the society—finance, manufacturing, education, communication, health, retailing, services—are growing at a steep rate, with no end in sight. This is unlikely to be a passing craze, but the enthusiasm could only be sustained by further evolution of AI that overcomes many of its current limitations. Propelled largely by the emergence of new machine learning algorithms and the availability of large labeled data sets and high computing power, AI programs for well-defined tasks such as recognizing faces or voices and playing games have surpassed humans in both speed and accuracy. Yet the human brain is capable of multi-tasking, learning with minimal supervision and generalizing learned skills, all accomplished with high efficiency and low energy cost. Most importantly, human brains can communicate with one another effectively via natural language—an ability acquired during early childhood within a communicative human environment. Understanding of syntax and semantics of human language remains one of the biggest hurdles in developing AI with human-like intelligence, ever since Alan Turing proposed the famous test of a computer's intelligence. If the ultimate goal of AI is to achieve human intelligence, then it would be useful to introduce structural and operational principles of the brain into the design of computing algorithms and devices. We are only at the beginning of understanding how the brain performs complex cognitive tasks—how we make decisions based on the accumulation of evidence, how we form concepts about the world, how we understand what others are thinking and think logically ourselves, and how we generate creative ideas. However, we do know very well how neural circuits are formed, how many neural systems are organized and how they process, store and retrieve information. In fact, neuroscience concepts have already been introduced into machine learning algorithms and played critical roles in triggering several bursts of advances in AI in past decades. In particular, the concept of synaptic plasticity, in which the weight of connections of the neural network could be modified, based on the network activity pattern or on the error of the output relative to that expected, has become the cornerstone of most machine learning algorithms. The goal of brain-inspired AI is thus to introduce more concepts and principles discovered in brain science into building AI computing systems. As the world's most populous country and the second largest economy, and having the highest number of internet users, China is now among the leading countries in commercial applications of AI technologies. However, very little pioneering work in AI technology development in the past was done in China, and current AI research in China remains largely focused on the application or incremental modification of existing technologies. The Chinese government has approved the funding of a major 2030 Science and Technology Project on ‘Brain Science and Brain-Inspired Technology’ (also known as ‘China Brain Project’), aiming in part to apply neuroscience knowledge for the development of next-generation AI with human-like intelligence and brain–machine interface technology. Thus we expect an accelerated development of brain-inspired AI technology in China in the coming decade. The march towards brain-inspired AI will also bring into focus a critical issue of the future AI: will machines with human intelligence be employed to perform work that most of us do for a living? Given the present status of AI technologies—even simple human cognitive processes could not be replicated by machines—human-like robots are still in the realm of science fiction. Nevertheless, AI systems with limited capabilities can and have already replaced many human jobs involving well-defined routines in manufacturing or data management. We could imagine in the coming decades the scenario of massive unemployment, resulting from the use of efficient AI systems and the lack of replacement jobs from those unemployed. President Xi Jinping pointed out in his congratulatory letter to the World AI Conference 2018 that ‘as AI is injecting new energy into the economic and social development … there is the need to satisfactorily deal with issues raised by AI technologies, including those in the areas of legality, safety, ethics, and governmental management’. The change in the social landscape is coming, but few societies are prepared for it. Given the distinct composition and educational level of the work force, different cultural and religious backgrounds, and varying structures of social welfare systems, advanced and developing countries have to deal with AI-triggered social and economic changes via different approaches. Discussion on the ethical and social impacts of AI technology is imminent, and consensus at the international level needs to be reached with full consideration of issues unique to individual countries.
No takes yet. Share an insight, caveat, or question.
Mu‐ming Poo (2018) studied this question.