![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Entry tags:
(no subject)
six month-ish diary update on fidding with AI at the dayjob (i briefly fiddle with new tools like this so i can base my own opinions on lived experiences versus third-handed anecdotes). so far the two use-cases remain limited to:
* "extend background" tool in photoshop. i don't use this but once every other month, but it's useful when a sales agent sends me a very badly cropped picture that i need to photoshop a rendered display onto, particularly when they don't give me a good amount of "floor" or retail store aisle to anchor it on. I honestly can't help but think this is at least pretty similar to photoshop's old clone too that's been around forever just on a larger scale. personally inoffensive.
* chat gpt strictly for (non-immediate) coworker birthday cards, with human editing afterwards. i happen to be the person that runs the birthday card hooplah for our department and i gotta tell you: i'm not writing individual personal cards for ~50 people every year, half who i've never met in person, though people who i talk to more than once every six months (and thus have some sort of conversation basis to write something personal off on) get a human-written one. personally inoffensive but i can see how somebody might be bothered if they knew.
* does photoshop's clone tool / various brush stabilizers across paint platforms count? i have a feeling a lot of people actually underestimate how common machine learning was even before the whole hooplah.
that's about it. pretty conservative list i think.
i also had a thought the other day i find most 3D animation/video game walk cycles so incredibly jarring to reality because it's not randomized enough (most people don't vaguely float in a repeated and predictable 2 second cycle when talking) and i wonder if there was a way machine learning could augment that for the sheer number of random edge cases, mixed with individual personalities. making a note to check into this.
* "extend background" tool in photoshop. i don't use this but once every other month, but it's useful when a sales agent sends me a very badly cropped picture that i need to photoshop a rendered display onto, particularly when they don't give me a good amount of "floor" or retail store aisle to anchor it on. I honestly can't help but think this is at least pretty similar to photoshop's old clone too that's been around forever just on a larger scale. personally inoffensive.
* chat gpt strictly for (non-immediate) coworker birthday cards, with human editing afterwards. i happen to be the person that runs the birthday card hooplah for our department and i gotta tell you: i'm not writing individual personal cards for ~50 people every year, half who i've never met in person, though people who i talk to more than once every six months (and thus have some sort of conversation basis to write something personal off on) get a human-written one. personally inoffensive but i can see how somebody might be bothered if they knew.
* does photoshop's clone tool / various brush stabilizers across paint platforms count? i have a feeling a lot of people actually underestimate how common machine learning was even before the whole hooplah.
that's about it. pretty conservative list i think.
i also had a thought the other day i find most 3D animation/video game walk cycles so incredibly jarring to reality because it's not randomized enough (most people don't vaguely float in a repeated and predictable 2 second cycle when talking) and i wonder if there was a way machine learning could augment that for the sheer number of random edge cases, mixed with individual personalities. making a note to check into this.