Continuing to play around with image uploading so it is more dependable - here is a sample of 25 random images from my ssd - of the 25 I asked it to upload in sequence initially, two partially failed, and eight failed to materialize altogether, and the node express api did not crash, which is a huge improvement.
This was accomplished with adding only one async call in the image upload loop to the angular side, but I have a feeling that continuing to tweek the node api will also potentially alleviate more issues. Regardless, errors can happen, and what with no proper error handling on the server, I should not be surprised my results are unsatisfactory. Some work has been done on the node side, but it’s mainly trial and error as I am a JS idiot.
Anyways - that’s one small improvement on the backend down - hopefully many more to go.
A photo of a punk on a skateboard stealing an old man's dog. A renaissance statue of a modern woman holding a living bird. A circuit board that looks like an overhead view of a city. A painting of a robot eating a banana in the style of Jean Giraud. A realistic photo of the end of the world as viewed from outer space by an astronaut. Photo of an old computer in the sand at the beach during a sunset with a skull next to it. A computer circuit board with red hair growing from the microchips. A photo of Michelangelo's Sistine Chapel but Adam is replaced with an owl and god is replaced with a spaceship.
Obviously, these are prompts to ask openAI to generate images via some weirdly trained machine learning model that they’ll further train by finding out which image I asked the model to iterate. Regardless, these are all hideous and they demonstrate the depths of the Uncanny Valley we currently find ourselves in. Can’t say I’m getting as impressive or as fun results from the MidJourney Discord, but I actually still have access to openAI, so this will have to do for now.
The other thing to mention is all the recent controversy about AI art as it is nothing more than a passable image created by averaging out meta data and images from other, real, human, content. I guess this is the end result of the internet - a bunch of ugly garbage created by non-artists to put artists out of work. Not to say all the AI art I’ve seen is bad, but judging by what people chose to hang on their walls and have tattooed on their bodies, the stuff generated by the black box will make people happy. And it t will only gain in popularity as people’s desires to pay artists disappears in place of a subscription to a learning model that makes you the artist instead (on the backs of millions and millions sets of data mined from the hours of work from real people).
The artistry of the future will be how well you can write a prompt.
Busy last couple days at work so here’s a sketch of some dude I did between coding sessions. Site also recently updated to use SSR but I have not quite figured out how to use it properly. For the time being, the server config is in place and I am using a ssr dist for now.
Made some backend updates to the site recently but its still not working correctly. Because the node server and angular app are on the same server, I had to set up a reverse proxy in nginx for it to properly address the node api I created. I don’t think I understand or have CORS set up well at all. At. All.
In the meantime, enjoy a desktop sketch I did while trying to think about how to solve a problem at work. It’s not bad for a five minute job! IMO!