EUVIMEDEUVIMED entdecken
ÜbersichtsarbeitOpen Access · CC BY 4.0

Brain-inspired strategies for efficient artificial intelligence

Min-Seo Kim, Hyoung F Kim

Molecules and Cells · 2026-07 · 49 (7) · 100365

DOI: 10.1016/j.mocell.2026.100365 ↗

Kurz gefasst

Übersichtsarbeit in der Zeitschrift Molecules and Cells, 2026-07. Verfasst von Min-Seo Kim und 1 weiteren Autor:innen. Beteiligte Einrichtung: School of Biological Sciences, College of Natural Sciences, Seoul National University (SNU), Seoul 08826, Republic of Korea; Interdisciplinary Progra…. Hauptthemen laut MEDLINE: Artificial Intelligence, Brain. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 87 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Research Foundation, National Research Foundation of Korea.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Despite being constrained by a rigid skull and a fixed number of neurons, the brain excels in processing diverse information and making adaptive decisions with remarkable efficiency. This raises a central question at the intersection of neuroscience and artificial intelligence (AI): How does the brain achieve such high performance with limited physical resources? We addressed this question by examining the anatomical funneling architecture of the basal ganglia, a central hub for value-based decision-making. Through parallel processing, distinct circuits support cognitive flexibility and habitual stability, enabling efficient allocation of neural resources and context-sensitive engagement in specialized computations. In contrast, convergent processing compresses the input across circuits, allowing the efficient extraction of core information for decision-making, a form of quantitative efficiency that minimizes the number of neurons required. However, this compression can degrade the fidelity. To address this, the brain employs qualitative efficiency in which population-level neural patterns preserve fine-grained information and support generalization across similar contexts. Finally, we propose that the cortico-basal ganglia system achieves cognitive efficiency by funneling the anatomical structure and dimensionality reduction to optimize both performance and energy demands. These principles offer a biologically grounded framework for developing brain-inspired, resource-efficient artificial intelligence systems that balance generalization with precision.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Artificial IntelligenceBrainHumansAnimalsNeurons

Stichwörter der Autor:innen

Brain efficiencyBrain-inspired artificial intelligenceDimensionality reductionFunneling architectureValue-based decision-making

Autor:innen und Institutionen

  • School of Biological Sciences, College of Natural Sciences, Seoul National University (SNU), Seoul 08826, Republic of Korea; Interdisciplinary Program in Neuroscience, Seoul National University (SNU), Seoul 08826, Repub…

Förderung

  • National Research Foundation
  • National Research Foundation of Korea

Zitieren

APA

Kim, M. S., & Kim, H. F. (2026). Brain-inspired strategies for efficient artificial intelligence. Molecules and Cells, 49(7), 100365. https://doi.org/10.1016/j.mocell.2026.100365

Vancouver

Kim MS, Kim HF. Brain-inspired strategies for efficient artificial intelligence. Molecules and Cells. 2026;49(7):100365. doi:10.1016/j.mocell.2026.100365

BibTeX
@article{Kim2026,
  author = {Kim, Min-Seo and Kim, Hyoung F},
  title = {Brain-inspired strategies for efficient artificial intelligence},
  journal = {Molecules and Cells},
  year = {2026},
  volume = {49},
  number = {7},
  pages = {100365},
  doi = {10.1016/j.mocell.2026.100365}
}
Kostenlos starten →