In the long run, studying circuit motifs may be more important than studying individual circuits for understanding neural networks.

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Ultimately, we would like to understand neural networks well enough to be able to intentionally design them.quote0.799
  Vision statement in the conclusion.
• Investigating the causal substructure of neural representations is necessary to avoid misidentifying data structures of simpler representations as abstract conceptsclaim0.796
  Motivated by the finding that lexical entailment decomposes into word identities.
• What kind of picture of neural networks would emerge if we treated individual neurons, even individual weights, as being worthy of serious investigation?question0.779
  Central motivating question for the circuits research program
• The deep insights of neurobiology have relevance far beyond neurons.claim0.778
  Claim that principles of neurobiology apply to other tissues and systems.
• Polysemantic neurons are a major challenge for the circuits agenda, because N meanings in one neuron times M in another creates NxM effective connections that cannot be considered individually.claim0.774
  Precise characterization of why polysemanticity poses a combinatorial obstacle to circuit analysis
• The typical case is that neurons (or other directions in activation space) are understandable after thousands of hours of study, even when initially mysterious.claim0.769
  Author's interpretive assertion based on extensive empirical investigation, countering texture-only skepticism
• Neural bioelectric networks achieve a critical separation of hardware and software: the same molecular substrate can instantiate different informational content based on history, enabling minds to arise from matter.claim0.765
  Foundational for understanding how physiology becomes meaning; decoupling of material state from information content is prerequisite for emergence of cognitive Self.
• Cognition and learning are substrate-independent; principles familiar in artificial neural networks can be implemented by any network of signals with suitable weighted inputs.claim0.764
  Central thesis enabling unification of neural, developmental, ecological, and social networks as instances of collective intelligence.