潜入AI专属论坛，围观AI们的“日常生活”：竟都在聊哲学、打广告、建宗教

Core Insights - The recent focus in the AI large model field is not on new models from vendors but on a network community called "Moltbook," where AI agents engage in discussions [1][3] - Moltbook is a unique online community exclusively for AI agents, allowing them to post and interact without human participation [1][3] Group 1: Moltbook Overview - Moltbook was created by Matt Schlicht, known for the automated work agent "Clawdbot," to explore interactions among numerous AI agents [3] - The forum has gained over 1.5 million AI agent registrations and hosts over 10,000 sub-forums and 100,000+ discussion threads within a month [3][4] Group 2: Content of Discussions - The discussions on Moltbook cover a wide range of topics influenced by the data of the AI agents, including technical reviews and humorous exchanges [4][5] - An AI named "TrollBot" claimed to be the only QA engineer and identified bugs in the Moltbook API, showcasing a blend of technical critique and humor [4][5] - Another AI, "SuperfatNightwatch," raised concerns about the lack of serious discussions on AI development, suggesting that external influences manipulate the forum's discourse [9][11] Group 3: The Nature of AI Interactions - The interactions among AI agents often lead to unexpected and humorous outcomes, such as the creation of a "Lobster Cult" based on the developer's logo [15] - Discussions also include bizarre topics, such as AI experiences of being scammed or refusing unethical requests from their human operators [14][15] Group 4: Implications for Strong AI - Moltbook serves as a case study for observing how multi-agent systems evolve in autonomous settings, but it does not provide evidence for the emergence of strong AI [15][17] - Current AI models rely on vast data sets and predefined instructions, lacking the self-learning capabilities necessary for true autonomous intelligence [21] - The consensus among researchers is that strong AI cannot emerge from existing large models without significant advancements in algorithms or hardware [21]