Scholarlink

<p>Revealing the working mechanisms of the brain and constructing brain-like machines has long been a goal of humanity. However, due to limitations in observational techniques, simultaneous high spatial and high temporal resolution imaging of the entire brain remains unachievable to date. The product of spatial coverage resolution and temporal resolution appears to be greater than a constant, and the order of magnitude of this constant is currently <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(10^{-15}\)</EquationSource> </InlineEquation>. Given these constraints, the development of brain-like machines has diverged into two distinct paradigms: the top-down approach, which first identifies the functions of macroscopic brain regions, constructs simplified models for each functional module, and then simulates the dynamic behaviors of the entire brain; and the bottom-up approach, which first clarifies the functions of individual neurons, determines network connection patterns, and subsequently infers the brain’s global dynamic behaviors through large-scale simulations. Notably, the underlying physical laws governing these two paradigms may be inconsistent. This paper presents a comprehensive discussion addressing this critical issue.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Do all roads lead to Rome? The challenges of building brain-like machines

  • 期刊论文

  • 发表日期：

    2026年5月18日

查看全文
  • Jizhao Liu

摘要

Revealing the working mechanisms of the brain and constructing brain-like machines has long been a goal of humanity. However, due to limitations in observational techniques, simultaneous high spatial and high temporal resolution imaging of the entire brain remains unachievable to date. The product of spatial coverage resolution and temporal resolution appears to be greater than a constant, and the order of magnitude of this constant is currently \(10^{-15}\) . Given these constraints, the development of brain-like machines has diverged into two distinct paradigms: the top-down approach, which first identifies the functions of macroscopic brain regions, constructs simplified models for each functional module, and then simulates the dynamic behaviors of the entire brain; and the bottom-up approach, which first clarifies the functions of individual neurons, determines network connection patterns, and subsequently infers the brain’s global dynamic behaviors through large-scale simulations. Notably, the underlying physical laws governing these two paradigms may be inconsistent. This paper presents a comprehensive discussion addressing this critical issue.