{
  "id": 16557096,
  "title": "Commentary: Citation graphs may underestimate neuroscience’s influence on AI",
  "url": "https://urgent.news/2026/10/26/commentary-citation-graphs-may-underestimate-neurosciences-influence",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-26T04:02:27.000Z",
  "source": {
    "name": "The Transmitter",
    "slug": "the-transmitter",
    "url": "https://www.thetransmitter.org/neuroai/commentary-citation-graphs-may-underestimate-neurosciences-influence-on-ai/"
  },
  "original_language": "en",
  "account": "Renato Duarte's citation analysis reveals the relationship between neuroscience and artificial intelligence, but it likely overstates the influence of neuroscience on AI. Simply put, citations do not effectively measure the intellectual debt a field owes to another. Modern AI is an engineering discipline, with a few conceptual breakthroughs spawning vast amounts of technical literature built upon them. As a result, the contribution of original neuroscience ideas can seem negligible, even if they significantly shaped AI's direction. The transformer architecture, which underpins modern large language models like ChatGPT, emerged from recurrent neural networks (RNNs) and the observation of recurrent neural circuits in neuroscience. Attention mechanisms were developed over decades in psychology and neuroscience before being applied mathematically in machine learning. When citing these concepts, citations naturally point to more recent machine-learning research, not the original neuroscience works. This \"obliteration by incorporation\" phenomenon leads to a systematic underestimation of deep, influential ideas in citation analysis. The citation habits of AI literature also reinforce this bias. AI papers tend to cite their most recent sources, with most references being 5 years old or younger. In contrast, neuroscience papers have a longer citation history, with a median age of 8 years. This means that the foundations of AI, such as the transformer architecture and the Hugging Face initiative's safety advances, often have unrecognized neuroscience roots. The most influential AI pioneers, like Demis Hassabis and Dario Amodei, were deeply involved with neuroscience during their doctoral research. Thus, while citation graphs provide a useful map of intellectual connections, they do not accurately capture the extent of neuroscience's influence on AI.",
  "summary": "An alternative, albeit less quantitative, way to trace this history is through the people who carried ideas between the fields.",
  "key_points": [
    "Citation graphs may overstate AI's influence, underestimating neuroscience's impact.",
    "Transformer architecture derived from RNNs and neuroscience's recurrent neural circuits.",
    "AI papers cite recent sources, hiding neuroscience's foundational contributions."
  ],
  "editors_take": "Reliance on recent citation data obscures the foundational role of neuroscience in artificial intelligence, as technical engineering papers frequently fail to credit the biological concepts that underpin modern architectures.",
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}