港大马毅外滩大会演讲：人工智能应从“黑箱”走向“白箱”

Core Insights - The current development of artificial intelligence (AI) lacks a scientific understanding of its essence, necessitating a shift from "black box" systems to "white box" models based on mathematical principles and closed-loop feedback [2][3] Group 1: Stages of Intelligence Evolution - The evolution of intelligence can be categorized into four stages: genetic intelligence represented by DNA, individual developmental intelligence formed by brains and perception systems, collective intelligence facilitated by language, and finally, true artificial intelligence [2] - Current AI, exemplified by large models, is still at the primitive "genetic intelligence" stage, relying on vast parameters and pre-trained data, which leads to high resource consumption and inefficiency [2] Group 2: Core of Intelligence - The essence of intelligence lies in the ability for "self-verification and self-correction," allowing for critical examination of existing knowledge to identify and rectify errors [2] - Current large models function merely as static repositories of knowledge, lacking true understanding, which results in logical confusion and "hallucination" issues [2] Group 3: Future Directions - The future of AI should focus on treating intelligence as a rigorous scientific and mathematical subject, aiming to develop systems with individual memory and autonomous closed-loop capabilities [3] - Progress towards true "autonomous intelligence" should be made within an explainable theoretical framework [3]