Biological brains have a high degree of precise internal structure that artificial neural networks lack.

Judge as of 2021-05-12 against the neuroscience and machine-learning literature. True: established evidence shows that biological brains have extensive, precisely specified internal structure (distinct cell types, specific connections and circuitry between regions, laid down by development and genes), and that standard artificial neural networks are comparatively uniform and generic, so this is a recognised difference. Mostly true: the contrast is broadly accepted, but 'precise' overstates it (e.g., much of brain wiring is stochastic or shaped by activity), or some artificial architectures have significant built-in structure. Mixed: experts are clearly divided on whether this is a key distinguishing feature. Mostly false: brains do have structure, but the evidence shows artificial networks of the time had comparable built-in structure, so the contrast is overstated. False: the evidence shows brains lack precise internal structure, or that artificial networks match them in this respect.

Set before the research began.

1. E1 A critique of pure learning and what artificial neural networks can learn from animal brains PubMed Central (NCBI), hosting a review article by Anthony Zador · 21 August 2019

   “Specifically, animals are born with highly structured brain connectivity, which enables them to learn very rapidly.”

2. E2 A critique of pure learning and what artificial neural networks can learn from animal brains PubMed Central (NCBI), hosting a review article by Anthony Zador · 21 August 2019

   “But in larger brains, such as those of mammals, synaptic connections cannot be specified so precisely; the genome simply does not have sufficient capacity to specify every connection explicitly.”

3. E3 Specificity and randomness: structure-function relationships in neural circuits PubMed Central (NCBI), hosting a review by Wei-Chung Allen Lee and R. Clay Reid · 18 August 2011

   “In both the retina and cerebral cortex, a great deal is known about topographic specificity, such as lamination, and cell-type specificity of connections. Little, however, is known about connections as they relate to function.”

Zador's 2019 review, which contrasts artificial networks with animal brains, says animals are born with highly structured brain connectivity (E1). Lee and Reid (2011) describe well-documented lamination and cell-type specificity of connections in the retina and cortex (E4). Together these support the claim that brains have extensive built-in internal structure. However, Zador also notes that in larger brains such as mammals', synaptic connections cannot be specified so precisely, because the genome does not have sufficient capacity to specify every connection explicitly (E2), so "precise" somewhat overstates it, which fits the criteria for mostly true.