Could you test ChatGPT vs MeaLeon TFIDF or HDBSCAN?
Could you devise a way to get ChatGPT to calculate a similarity metric? or can you use triangle inequality and ask in a roundabout way: is if I like chicken and brown rice, would I like more probably fish and chips or chocolate cake?
For the triangle inequality, we could try to have ChatGPT group by macros?
If I like chicken and brown rice, would I be more likely to like fish and chips or chocolate cake?
like pick two mealeon suggestions and two others and just ask cgpt it’s thoughts on these meals and similarities, and which ways it thinks would be most suitable for finding similar food suggestions but not same food?
"Art is how we decorate space, music is how we decorate time.” —Jean-Michel Basquiat
TODO Fix the lore aka quote timer bot (it isn’t displaying quotes)
TODO #MeaLeon give searcher the option for which ‘measure’ to use. In current state, TF-IDF probably produces more similar recipes, but the OHE produces wild suggestions which can be also interesting
LATER Maweexy requests late night ASMR stream voice
LATER integrate ChatGPT to ask what it thinks about similarities and differences
Maybe use it to suggest allergy or dietary workarounds
Could also ask "what do you think is similar and different between these two recipes"
When looking at the results it suggests, what is the difference between its suggestions and ‘mechanical Turk’ or a Google search? What is the level of sophistication?
LATER make a Chrome extension that knows your allergies (and risk tolerance) and just ask it "will this kill me?"