ai is so lame (5 reasons)
- the most popular models are trained by huge corporations, which means that the output will always be as inoffensive (and therefore, bland) as possible.
- every person who is saying that ai will replace developers has not sent more than 10 requests to any llm
- the training data is us-centric, so the responses you could get from a chat bot is an average redditor response. no offense to redditors, i go there sometimes as well (m'lady), but it is very very bias in the ways that are not helpful.
- your requests can be used to train models further. i know that in some paid plans, the open ai / antropic / etc are claiming to not train llms on your chats, but can you be really sure? my hot take: unless the model runs locally, no you can not. (i have tried to run some model with lm studio and my computer froze, honestly i might start looking into it a bit more)
- ai art looks very ai and lame. ai texts are so condescending, i cannot tell if they have always been this way or if this is a new development. the issue i'm having with it is not that it is "inhuman" (whatever that means), but that it is trashy. side note - i wonder if pixar would have to change their cartoon style? since a lot of llms have been trained on pixar data, their generated images remind me of pixar to the point that i am revolted by pixar.
talking to an llm is like taking to a representative of a mega corporation, who has taken what people have said on forums, filtered it through its bullshit pr filter, spit out the average of what millions of people have written, while trying to sound human with the emojis and whatnot. for real. there is nothing insightful there by design, nothing cutting edge, nothing emotional or fun. oh and chat gpt is gonna have ads now, no reason to think other llms won't. ugh. ugh!