Thanks for replying 🙏
Sure .. grifters gonna grift ... and if you're playing rug-pull games with over-hyped investments expect to get, unfortunately, rug ... pulled. Defo not a clown-world defender and agree that an aspect like that is a horrible hoax with undeserving victims.
But ... What are the other soultions before LLMs? ... If you program, what tools could I use in a method that takes (text, required_key_value_extracted_pairs) and returns a JSON {"key_1": "value_1", ...} object? Advanced LLMs, with guidance, really excel in that functionality. After all, they're just heavily trained neural networks ..
In particular, I have a "chat extract" method in my FOSS API, https://ajrlewis.com/api/docs, that uses a HuggingFace interference model (probs LLama, can't remember lol!) to do this task ... Would love to know a different approach if you can suggest one 😊 Defo save on the overheads + complexity if I didn't want to use a 3rd party (HuggingFace) in future ...
i have written many scrapers of various kinds over the years, tree walking syntax extracting things
i think you just have never bumped into the black art of language processing at all, LLMs do advance it in some respects but certainly not in energy efficiency or storage efficiency, nothing is better than a well designed language, and i'm in a minority in this opinion, the one composed of all the people who choose #golang instead of rust or c++ or javascript, all of which are incredibly complex, wasteful, and hard to read languages that also cost a lot of human training time for no reason except to comfort the small group of "expressivity" cultists who dominate corporate IT departments
also, as regards to writing stuff properly, i learned to touch type, yes i hate capital letters because pinkies are not made for pressing so much as germanic based english capitalisation requires, and fuck germany this is not their world anymore
Thread collapsed
