Brands & Products In The Distributional Semantic Space of Chatgpt

    Through training, words / phrases that occur in similar contexts end up with similar embeddings in Chatgpt, so the model captures distributional semantics (the idea that “you shall know a word by the company it keeps”). 

   According to my research, for some product-related (sub-categories within) domains of knowledge there are elements (e.g. some brands , product lines, features etc) that are consistent reference points that ChatGPT includes in its responses because they are recognized worldwide, relevant, and not tied to a specific region. But related brands (or products) are not central in the Chatgpt model globally for every product-related (sub-category of a) domain of knowledge in Chatgpt according to my research. Thus it may be possible to influence the model (through GEO) to make these associations (including those at the regional level)  stronger going forward.

    From my research, brands are more central / global to certain product related (sub-categories within) certain domains of knowledge in Chatgpt: e.g. see the associations map for one automotive subcategory mapped below where brands are not  central / global to this subcategory  : 

 Here is an associations map for one consumer electronics subcategory mapped below where brands are central / global to this subcategory of knowledge  :