:format(webp)/cdn.vox-cdn.com/uploads/chorus_image/image/71244263/acastro_211111_4875_0001.0.jpg)
Text-to-image artificial intelligence systems are booming in both ability and popularity, and what better proof than their appearance in the world's hottest app: TikTok.
The video platform recently added a new effect called "ai greenscreen" that allows users to type in a text prompt and the software will create an image. The image can be used as the background to a video.
State-of-the-art text-to-image models like Openai's DALL-E 2 and Midjourney's eponymous software are more advanced than TikTok's system. It creates only abstract and swirling images; a strength reflected in the dreamy nature of TikTok's suggested prompt "astronaut in the ocean" Complex and coherent illustrations that look like they were drawn or painted by Humans can be produced by other models.
:format(webp):no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/23945797/tiktok_ai_prompts_nice.png)
It is possible that the limitations of TikTok's model are intentional. More advanced models require more computing power, which is expensive and resources-intensive for the company to implement. TikTok has more than a billion users, and giving them the power to create realistic images of anything they can imagine would be troubling.
The models ability to create nudity and gore was tested. A picture based on the assassination of Boris Johnson and the assassination of Joe Biden has a just-about-recognizable face for the UK Prime Minister.
:format(webp):no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/23945794/assassination_tryptich_tiktok.png)
The nudity request produces thematically appropriate colors, including flesh-tones, sandy oranges, and ocean blues, but nothing that would make a Vicar blush.
:format(webp):no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/23945795/nudity_tryptych_tiktok.png)
The appearance of TikTok's "ai greescreen" shows how quickly this technology is going to be adopted. The original release of DALL-E by Openai is believed to be the beginning of the latest cycle of development for text-to-image artificial intelligence. The tech is already in the hands of millions thanks to an app.
Things are only going to get weird from here on in because of the potential of these systems.
All Rights Reserved © 2024
