我国科学家破解大脑灵活决策底层奥秘,单个神经元就好比一台微型计算机,或将启发下一代人工智能的底层算法。
中国科学院脑智卓越中心科研团队首次揭示大脑神经元参与灵活决策的计算机制,发现单个神经元如同微型计算机,其树突和胞体相互作用可完成认知计算,有望启发下一代人工智能底层算法。
AI-generated summary
中国科学院脑智卓越中心的科研团队通过实验首次揭示了大脑神经元参与灵活决策的计算机制。
全释硬科技丨神经元新发现或将重塑人工智能
我国科学家破解大脑灵活决策底层奥秘,单个神经元就好比一台微型计算机,或将启发下一代人工智能的底层算法。揭秘人脑为何比AI更灵活。
近日,中国科学院脑智卓越中心的科研团队,首次揭示了大脑神经元参与灵活决策的计算机制,填补了长期缺失的空白。这项成果,很可能启发下一代人工智能的底层算法。
AI和人脑的差距究竟在哪?大模型最擅长的语言,AI的图像识别,训练量至少都是人类的上万倍。同样是学习,为什么人脑能学得更快呢?答案就藏在大脑的基本结构——神经元里面。
大脑里有很多神经元,我们可以把它们想象成一个个小车站。车站和车站之间,靠“突触”连接,就好比车站之间的路。学习,本质就是让神经元之间的通路变得更加顺畅。
为了破解大脑灵活决策之谜,科学家设计了一套听觉抉择实验,训练小鼠玩猜规则游戏。小鼠会听到不同声音,需要找到隐藏的规律。等小鼠摸透当前规律后,科学家再悄悄更换这套规则,观察小鼠如何快速适应新逻辑。
科研团队利用双光子显微镜,观察小鼠听觉皮层的神经元。神经元并不是一个光秃秃的小圆点,它长出很多像树枝一样的突起,这就是树突,长长的那支远端分支叫作顶端树突。顶端树突负责承载推断出来的“抽象规则”,细胞本体也叫胞体、负责接收耳朵传来的感官信号。两套信息,在同一个细胞里就完成比对运算——相当于一边提出猜想,一边接收现实,当场判断猜想是否成立。
这个实验也就证明:单个神经元本身就是一台小型计算机,神经元的树突和胞体相互作用就能完成决策相关的认知计算。只有猜想经过多次验证、确认可靠之后,大脑才会慢慢加固突触,把这条规则永久保存下来。
现在的AI,只模拟了神经元之间的突触连接,也就是我们前面说的“车站之间的路”。如果能把这次新发现的神经元计算新机制融入算法,就有望打造出能快速提出假设、少量样本就能推理新规律的新一代人工智能。亿万年演化,把精巧的计算能力直接塞进人类大脑的单个神经元,我们现在,或许才刚刚揭开它奥秘的冰山一角。
AI outlook — possibilities, not facts
将新发现的神经元计算机制融入下一代人工智能底层算法
Likely · Within months
The Chongqing Science and Technology Awards Conference was held on the 16th. Professor Wang Shilong of Chongqing University won the "Outstanding Contribution Award in Science and Technology". Wang Shilong has long been committed to the research of gears and industrial machines. Through forward design and deep integration of industry, academia and research, he promotes the independent controllability and industrial upgrading of my country's high-end equipment manufacturing core technologies.

Chinese scientists have discovered a 150-million-year-old fossil of Yuezhuornis yangyuanensis in Fujian province, the earliest known flightless bird, challenging theories that flightlessness in birds evolved only as a secondary adaptation.
The Academy of Sciences of Developing Countries announced the results of the new council election on October 14, local time, and Hou Jianguo, President of the Chinese Academy of Sciences, was elected president of the Academy.
China's total R&D investment in 2025 will reach 3.92882 billion yuan, an increase of 8.2% over the previous year, ranking second in the world, second only to the United States. The growth rate of basic research funding was 13.3%, 5.1 percentage points higher than the overall growth rate. Enterprises have become the main force driving growth, accounting for 77.4% of the entire society's R&D funding.

The team of Chen Shenghong, associate researcher at the Institute of Molecular Biology, Academia Sinica, discovered that cells form a natural firewall that blocks the spread of death through their own arrangement. Cell density and direction determine the propagation path of the death wave. The research results were published in the international journal "Cell".

Ya Qi, winner of the 2025 Nobel Prize in Chemistry, gave a speech at National Taiwan University, introducing new materials such as metal-organic frameworks and covalent organic frameworks, which can absorb water vapor or carbon dioxide in the air and release water for collection through heating, providing new directions for drought resistance and climate change mitigation. He also emphasized the role of artificial intelligence in materials exploration and the "selective science" thinking that humans need to judge important issues and research directions.