Doing Things Myself
Dec. 1st, 2023 08:24 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
krpalmerOn the drive in to work yesterday, the radio program I listen to had pieces on ChatGPT on either side of the half-past local news. I heard just enough of them to look up the convenient reference of a Wikipedia article and confirm it had been one year since the “text-to-text” generator program had become available. That means it’s been a year of never quite getting around to trying it out myself.
Perhaps I write enough I’m not interested in having that done for me, but I also have to admit that at the time it became available I’d been dabbling in the “text-to-image” program Stable Diffusion. It didn’t take long until I was aware of all the indignation about these programs being fed vast quantities of material made by people, but it also might not have been long until “that it can be done at all” faded into an awareness of the fine details going wrong. “Okay, surprise me” turned into a sense I’d come to the limits of the model files I had.
Some months after my interest faded I did start the program again, and then I got annoyed enough with the things going wrong that I picked up the “Apple Pencil” I’d bought some time ago and started drawing using my iPad. I may even have drawn enough since then to have the sense of identifying certain problems and details alike and improving a tiny amount, but I’m still stuck with the thought I’m not deft enough yet at what I’m trying to go ahead and prove what I’m claiming with mere words using pictures. In any case I can find myself thinking that when it comes to the risks some are very intent on identifying in “artificial intelligence,” simply deciding to follow its apparent answers might be a problem too.
Perhaps I write enough I’m not interested in having that done for me, but I also have to admit that at the time it became available I’d been dabbling in the “text-to-image” program Stable Diffusion. It didn’t take long until I was aware of all the indignation about these programs being fed vast quantities of material made by people, but it also might not have been long until “that it can be done at all” faded into an awareness of the fine details going wrong. “Okay, surprise me” turned into a sense I’d come to the limits of the model files I had.
Some months after my interest faded I did start the program again, and then I got annoyed enough with the things going wrong that I picked up the “Apple Pencil” I’d bought some time ago and started drawing using my iPad. I may even have drawn enough since then to have the sense of identifying certain problems and details alike and improving a tiny amount, but I’m still stuck with the thought I’m not deft enough yet at what I’m trying to go ahead and prove what I’m claiming with mere words using pictures. In any case I can find myself thinking that when it comes to the risks some are very intent on identifying in “artificial intelligence,” simply deciding to follow its apparent answers might be a problem too.