Nature重磅：图灵预言的AGI早已实现，人类却不敢承认

Core Viewpoint - The article asserts that Artificial General Intelligence (AGI) has already arrived, with large language models like GPT demonstrating human-level intelligence [1][3]. Group 1: AGI Definition and Evidence - AGI is no longer a distant dream; it is present and observable through everyday AI tools [3]. - The research team from UCSD argues that AI is not just intelligent but has achieved true generality, surpassing human capabilities in various tasks [7][8]. - The article compares this realization to historical cognitive revolutions, suggesting that AGI challenges human-centric views of intelligence [8][45]. Group 2: Capabilities of Current AI - Current large language models (LLMs) exhibit a breadth of capabilities across mathematics, language, science, and creativity, aligning with the definition of general intelligence [12][16]. - The models have demonstrated performance in international competitions and can assist in advanced research, indicating a level of intelligence comparable to that of humans [12][16]. Group 3: Misconceptions and Resistance - The article highlights a disconnect between the capabilities of AI and the skepticism from top AI researchers, with 76% believing current methods are unlikely to achieve AGI [12]. - The authors argue that this skepticism stems from a combination of vague definitions, emotional fears, and commercial interests [13][41]. - The paper systematically addresses ten common objections to AGI, demonstrating that current AI can solve novel problems and engage in complex reasoning [35][40]. Group 4: Historical Context and Future Implications - The article draws parallels between the current state of AI and historical shifts in human understanding, such as those brought by Copernicus and Darwin [45][46]. - It suggests that the emergence of AGI will force humanity to redefine its understanding of intelligence and its own place in the world [46][47]. - The discussion emphasizes that recognizing non-human forms of intelligence is crucial for future interactions with AI [47][48].