In some sense, this is the simplest language model we profoundly don't understand. And so it makes a natural target for our paper.

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

• Language models are some of the most remarkable computer programs in existence.quote0.819
  Opening sentence setting the stage for the importance of interpretability.
• Language models are few-shot learners (Brown et al., 2020)concept0.819
  Demonstrated transformers on mathematical understanding and logic; cited to motivate transformer versatility.
• The surface syntax of natural language don't offer much.claim0.814
  Claim about the limited utility of natural language surface features.
• It's tricky, because for a typical language model the entity is sort of tricameral: the base simulator, the simulated simulator, and the simulated awareness.quote0.808
  Antra's earlier definitive statement of the tricameral model.
• Today's Large Language Models have become so good at playing Turing's game that it often takes experts to demonstrate the present limits of their ability to simulate human-like intelligence.claim0.795
  Paper's assessment of current LLM capabilities relative to Turing Test
• Large language models develop surprisingly coherent yet often rigid internal preferences as they scalefinding0.793
  Mazeika et al. finding reinforcing the need for emptiness-based flexible value architectures
• That such a simple form-language could produce such powerful results, was remarkable.claim0.792
  Reflection on the eleven principles class, affirming that even a minimal form language can yield strong results.
• We shall only be able to reproduce versions and combinations of what can be 'reached' by that form language.quote0.792
  States that the form language delimits what buildings can be created.