this is super valid and great to come to as an epiphany in a blog post if you're a college kid or a random schmuck and kinda weird to do if you've held a high prestige AI job for a full year.
this is super valid and great to come to as an epiphany in a blog post if you're a college kid or a random schmuck and kinda weird to do if you've held a high prestige AI job for a full year.
It's at least interesting in the meta sense of noting that the pedagogy of how people are learning or self-teaching AI stuff appears to be deviating from machine learning / stats fundamentals.
@ryxcommar There is something funny with OpenAI… I’ve noticed a lot of these kind of cracks in the armor that don’t really make sense. Even the way Altman sometimes speaks… I mean the product speaks for itself but I guess I’m just a little surprised that ChatGPT is so good
from a regulator’s perspective, it’s so bleak and v concerning that the people allegedly leading an industry (& whose actions will dictate how a thousand companies will structure their business models, appeal to VCs, advocate to lawmakers) broadly do not understand the most elementary components of what they’re doing
@ryxcommar don’t really think this is trivial or obvious by any stretch
Yeah, „garbage in garbage out“ is kinda the first thing you learn. And i don’t believe its not something you realise after looking into model behaviour and underperforming data points for 5 minutes in your first model. Maybe they just wanted to make content and this was a low hanging fruit?
@ryxcommar im starting a cult of claude shannon until people re-learn stats and information theory
@ryxcommar Gonna walk around with a huge mirror and call it an optimized model to simulate my surroundings