Do AI models reason or regurgitate?

"Many people reacted defensively, insisting that today's AI systems are nothing more than "stochastic parrots" that regurgitate memorized information and are structurally incapable of emergent reasoning. I appreciate that many people want this to be true, but it is an outdated narrative based on a 2021 paper that was widely misinterpreted. I want to clear up this misconception because I worry it is giving the public a false sense of security that superintelligence is not an urgent risk to society."

"Like it or not, there is increasing evidence that frontier AI systems do not just store text patterns as statistical correlations, but also build structured internal representations (i.e., "world models") that abstract the concepts described in the ingested content. This is not surprising since human brains do the same thing - we consume information as language but store it as abstract conceptual models."

"One early study on this front showed that large language models (LLMs) trained only on language describing the rules of a common board game (Othello) contained an internal map of the board state (i.e., a world model). Researchers were able to "edit" the internal model and change the state of pieces on the board. This suggests a conceptual encoding of information that goes far beyond stochastic language correlations."

Public belief that current AI systems are only "stochastic parrots" underestimates emerging capabilities and risks. Frontier AI systems increasingly form structured internal representations or "world models" that abstract concepts from text. Human brains similarly convert language into conceptual models, and although brains are more efficient, large AI systems are demonstrating conceptual encoding. Experiments show LLMs trained on game rules can contain editable internal board-state maps, and training on descriptions of places and events can produce representations of location and time. These findings imply AI capabilities exceed simple statistical pattern matching and are rapidly improving.

Read at Big Think