Toward bridging the gap between machine intelligence and machine wisdom: Dilemmas and conjectures

In recent years, artificial intelligence (AI) has achieved tremendous development, akin to a significant leap, similar to progressing from 1 to 100. However, a significant gap still exists between current machine intelligence and human wisdom: machine intelligence is constrained to post hoc inference based on existing data, lacking the ability for genuine exploratory innovation and possessing no prospective reasoning inherent to human wisdom. Drawing inspiration from human wisdom, this article presents conjectures for overcoming the four dilemmas faced by machine intelligence: neglect of silicon-based cognition, lack of artistry, pitfall of perfectionism, and obsession with uniformity. These conjectures aim to propel machine intelligence toward machine wisdom, achieving a great leap from 1 to i.

Introduction

The evolution of machine intelligence has followed a non-linear trajectory, characterized by significant setbacks and periods of stagnation, often referred to as artificial intelligence (AI) winters. These challenges have complicated the progression from foundational research to advanced applications. However, recent breakthroughs in deep reinforcement learning1 and large language models have substantially advanced the field, enabling the integration of AI into diverse areas of human life2 and, in some cases, exceeding human-level performance.

However, the evolution of machine intelligence faces inherent limitations, as a significant gap remains between AI and human wisdom. As highlighted by Oxford researchers,4 AI relies predominantly on pattern recognition and data-driven learning, restricting it to replicating existing knowledge rather than generating novel insights. Unlike human cognition, which advances through hypothesis formulation and empirical verification, machine intelligence lacks imagination and is confined to post hoc inferences based on available data, failing to achieve prospective reasoning.

The critical question remains: how can we evolve machine intelligence into machine wisdom? By examining the genesis and development of human cognition, as Figure 1 shows, we identify four principal challenges that differentiate contemporary machine intelligence from human wisdom: the neglect of silicon-based cognition, the lack of artistry, the trap of perfection, and the obsession with uniformity. Addressing these limitations is essential for advancing machine intelligence to its next evolutionary phase, ultimately achieving a transformative leap.