Under Chantal, there use to be Word2Vec, now replaced by FastText. Aka a shallow neural network building a language model that understand synonyms and some EN/FR translations, by embedding words as vectors in a vector space built so cosine distance represent semantic distance. It was called AI when Google published Word2Vec and Facebook FastText, I didn't made it up.
I mean, I wrote 2 articles on the topic explaining how it's built. Each page in the index is represented by the centroid of all the word vectors it contains, then it computes the cosine distance between the centroids of the pages and the centroid of the request and returns the closest matches. Aka not your regular grep.
Overly dramatic yourself, Internet minion. All opinions and zero substance.
Please calm down, I had no intention of offending you! I agree, of course that's AI, just a pretty... commoditized? kind in 2026 and a long-winded explanation of word2vec or even mention of it is probably not necessary. What I meant it's not the sort of model anti-AI people would oppose, in response to an "instant pass" knee-jerk reaction.
Under Chantal, there use to be Word2Vec, now replaced by FastText. Aka a shallow neural network building a language model that understand synonyms and some EN/FR translations, by embedding words as vectors in a vector space built so cosine distance represent semantic distance. It was called AI when Google published Word2Vec and Facebook FastText, I didn't made it up.
I mean, I wrote 2 articles on the topic explaining how it's built. Each page in the index is represented by the centroid of all the word vectors it contains, then it computes the cosine distance between the centroids of the pages and the centroid of the request and returns the closest matches. Aka not your regular grep.
Overly dramatic yourself, Internet minion. All opinions and zero substance.