Hip-Hop Sampling vs. Scraping Data for Art
“When a text prompt comes to life as you envisioned, it feels a bit like magic. When technologies feel like magic, adoption rates tick up rapidly.” — Charlie Warzel
The above image was created using Midjourney AI, a proprietary artificial intelligence program that creates images from textual descriptions, similar to OpenAI’s DALL-E and the open-source Stable Diffusion. It works by scraping the Web and swapping text for pixels from AI ‘bots’ trained on thousands of images.
Once it became ‘all-the-rage’ the critics came out of the woodwork but I thought the process of making images using Midjourney and other AI tools like it reminded me of sampling in hip-hop. More specifically, I was reminded of “All Samples Cleared,” an album by the late Biz Markie. A sampler is an electronic device that records, alters, and plays back digital audio. It employs digital samples of real instruments, snippets of songs, or sound effects to create music.
Biz sampled a portion of the song “Alone Again (Naturally)” by singer-songwriter Gilbert O’Sullivan in the track “Alone Again” from Markie’s third album, I Need a Haircut. Markie and his production and recording companies were listed as co-defendants with Warner Bros. in a lawsuit.
The court case had a major effect on hip hop music. Sample clearance fees prohibited the use of more than one or two samples for most recordings, with some mechanical rights holders demanding up to 100% of royalties. As each sample had to be cleared to avoid legal action, records such as those produced by the Bomb Squad for Public Enemy, which use dozens of samples, became prohibitively expensive to produce. — Wikipedia
Rather than scrape a large dataset of images, the digital sampler grabs snippets of a set number of tracks (one song at a time). When I was enrolled in a SAIC Graduate Studies class with professor Shawn Decker, I learned how to use a sampler (machine) to produce my own music tracks. I recall sampling heavily sampled songs such as “I Know You Got Soul” by Eric B. and Rakim. The result was like a sonic collage.
NSynth Super is an AI equivalent to the digital sampler and it generates completely new sounds. On the flip side of this process is Midjourney. The images generated from any number of text prompts can (like NSynth) be unique: a completely original composition made up of snippets of images.
The image above was not generated using one photo of Viola Davis, rather the AI bot scraped several images of her to create this one. Other words in my prompt created the visual elements that surround Viola. This is the visual equivalent of digital sampling (really good producers made it very difficult to identify samples in a track). Critics are asking, “But is it art?” and I remember them saying the same about 90s rap (“Is it music?”).