Nice article, and I like the choice of the word "integration" rather than "generalization" to describe the ability of a model to take an internal representation and apply it in a new scenario.<p>I continue to think that Relational Frame Theory [0] both explains <i>why</i> these models work so well, and also <i>how</i> they're able to integrate knowledge through nothing but language. I believe that a researcher could show that LLMs emergently encode "frames" that describe relationships between concepts; that frames can be combined to form more complex expressions; and that frames can be reused in different contexts to make "novel" connections.<p>[0]: <a href="https://en.wikipedia.org/wiki/Relational_frame_theory" rel="nofollow noreferrer">https://en.wikipedia.org/wiki/Relational_frame_theory</a>
The best way to think about LLMs since even before ChatGPT has always been the simulators essay <a href="https://generative.ink/posts/simulators/" rel="nofollow noreferrer">https://generative.ink/posts/simulators/</a>
Interesting "brain drop," as the author describes it. It reminds me, though, of theories of human language that treat language as cognitive processing inside individuals' minds rather than as a social phenomenon. It might be more useful if we think about LLMs in terms of how they interact with us and with other software.
The article claims to balance the views of the promoters and detractors of LLMs, but already in the title it uses terms like "digital mind" and "knowing" which will probably be strongly opposed by the detractors...