It is strange that this milestone of machine learning rarely gets discussed.
I have a feeling that your typical smart software engineer (probably works on large-scale infra) became increasingly jaded re: ML, while the field continues to deliver astonishing achievements regardless.
Maybe its the usual ressantiment towards "a hyped field", which only strengthened after the advances continued and the prophecied neo-AI winter failed to happen. Maybe its compensation envy. Maybe math envy. Maybe disdain for softer " faulty" error-prone computing paradigms in general.
Anyway, as a non-ML software engineer I find the field genuinely interesting from mathematical, engineering and philosophical POV. Recent advances such as CLIP became very practical even for small-scale personal use.
If you are interested in large multimodal models, feel free to chat.
VQGAN-CLIP is fun. Got it working on my GPU a couple weeks ago. You reminded me of this post from the other day, a link to the paper "Prompt Engineering for Text-based Generative Art". Thank you to whoever posted this :)
3 comments
[ 3.1 ms ] story [ 20.2 ms ] threadI have a feeling that your typical smart software engineer (probably works on large-scale infra) became increasingly jaded re: ML, while the field continues to deliver astonishing achievements regardless.
Maybe its the usual ressantiment towards "a hyped field", which only strengthened after the advances continued and the prophecied neo-AI winter failed to happen. Maybe its compensation envy. Maybe math envy. Maybe disdain for softer " faulty" error-prone computing paradigms in general.
Anyway, as a non-ML software engineer I find the field genuinely interesting from mathematical, engineering and philosophical POV. Recent advances such as CLIP became very practical even for small-scale personal use.
If you are interested in large multimodal models, feel free to chat.
https://news.ycombinator.com/item?id=31312882
Though generally artists and programmers tend to keep a few good tricks just for themselves.