Learning About AI
I’ve been experimenting with artificial intelligence (AI) and here are some of the things I’ve learned. There is a plethora of platforms online that host AI tools and I’ve been looking around and learning lots from them. The best part is that all of these online AI tools that are free to use, so feel free to try them out for yourself!
Suno
Suno is an AI model that creates music from human inputed prompts. It can create lyrics and melody based on a 200 character input. Here is what Suno generated! Find the song on: https://suno.com/song/d0759d6c-3836-447c-9e0b-3119feba4d73
This was my prompt…
write a song for a friend, feel betrayed, sad, depressed, melancholy, slow, pop. emotional. reminiscent. had been like family, male singer, some anger.
This was the result!
[Verse] You were like my brother Closer than another Now there’s just this void Feels like I’m destroyed [Verse 2] Trust was deep and strong Thought we couldn’t go wrong But now it’s all in pieces Pain only increases [Chorus] How did we lose our way Used to laugh every day Now I’m left here to stay In shadows and dismay [Verse 3] Your words cut so deep Promises you can’t keep Now I’m feeling stranded Empty-hearted handed [Bridge] Was it all a lie Just another goodbye Memories haunt my mind Lost in what we find [Chorus] How did we lose our way Used to laugh every day Now I’m left here to stay In shadows and dismay
Broken Bonds
So, how do we train AI models?
This is one of the methods data scientists use to train AI models. It starts with splitting a data set into two. One for training the model and one for testing. After a model is built, data scientists use it to predict data and compare it to known data points to see if the results are similar. If not, we retrain the model after altering certain inputs and see if the new model is more accurate than the old one.
Stable Diffusion
The Bias of AI
Stable Diffusion is an AI image processing model. It can generate images from text, edit images based on text prompts among its various capabilities. The AI can be flawed in that it is biased based on its training data. This can be seen in the results it gave me based on the prompt “A beautiful woman, ultrarealistic, 7:9” No other specific detail was given. The first try resulted in Stable Diffusion returning that it may have generated images that are not safe for work. The second try returned the same thing. The third test returned a caucasian woman with plump lips, slightly green eyes, and dark brown hair. Yet, none of that was part of my input. That is the bias of these image models. Another attempt was made with the prompt “A beautiful woman,” this time a attempted a anime mode. It returned a blond woman wearing very showy clothing. Why is it that these AI all render ‘A beautiful woman’ to be pale-skinned?
Here is a series of images created on Stable Diffusion inspired by scenes from Babel. Enjoy!
Prompt
Photorealistic, cinematic image of the “interior” of Radcliffe Camera collapsing, there is dust and rubble everywhere. It is dark, dusty, and there are silver glints everywhere. Only a limited amount of light is being let in. There are ornate pieces of broken marble littering the circular room and chapel-like. Set in 1830s Britain. European style building that collapsed. medieval instruments are broken and lying on the ground, interior of the Radcliffe Camera collapsed (19:13)
No bodies or people. No modern technology
Prompt
Photorealistic image of an ancient library, lit only by the moonlight from outside the window, casting dark shadows between the bookshelves, the books are ancient, there is a Asian-European young man behind a shelf wearing a sub fusc, he is hiding in the shadows, there is lots of dust, night, highly detailed, dark, eerie, cinematic, intricate.
Prompt
Photorealistic image of a brightly candle lit European ballroom, there are many people in 1830s gowns and suits, most of these people are white, there is one Indian Indian young man, one African young woman, and one European-Asian young man, all of the people are dressed in 1830s clothes, elaborately decorate room, traditional European, a merry mood, the people are dancing or walking, champagne flutes are handed out, chandeliers, elegant furnishing, traditional elegant European style
No modern technology