It has long been said that programming languages need more of the features of natural language, but it has not been clear what the desirable features are.

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

• A programming language is low level when its programs require attention to the irrelevant.concept0.819
  Epigraph by Alan Perlis motivating the search for abstract conceptual models in Fruit's design philosophy.
• The surface syntax of natural language don't offer much.claim0.819
  Claim about the limited utility of natural language surface features.
• Studying programming systems requires a paradigm shift from programming languages, where interaction between programmer and system becomes central rather than program code alone.claim0.797
  Core assertion that systems perspective is incommensurable with language perspective; interaction, not code, is what matters in systems analysis.
• Language models are some of the most remarkable computer programs in existence.quote0.790
  Opening sentence setting the stage for the importance of interpretability.
• Pure functional languages cannot be the whole story for parallel programming because they fail to provide needed expressivity.claim0.785
  Argument that recursion equations are inappropriate for many important parallel programs.
• It is impossible for us to accept the claim that functional languages are necessarily higher-level than C-Linda.claim0.782
  Critique of the functional-level argument based on mental models.
• a basic difficulty [of non-functional languages] is that they are very long winded, in terms of the amount one has to write to achieve a given effect.quote0.776
  From Turner (1982), cited to represent the functional programming community's view and then refuted.
• In some sense, this is the simplest language model we profoundly don't understand. And so it makes a natural target for our paper.quote0.771
  Articulates why a one-layer transformer with MLP is the appropriate starting target for mechanistic interpretability