For a long time, humans have dreamed of machines capable of effortlessly generating art instantly, and this dream is becoming a reality in recent years. New deep-learning AI models are being developed to create pieces of art based on a prompt consisting of a few words. The process by which it creates the images is very complicated and time-consuming to learn but to put it in simpler terms, it uses a large set of images and captions to make predictions on what the prompt would look like. Then it creates multiple possible images using its algorithm.
Throughout 2022, many new AI Art text-to-image models got announced: In March 2022, MidJourney was available in a Discord server; In April 2022, OpenAI announced DALL-E 2; and in August 2022, Stability AI released its open-source image generator Stable Diffusion. These machine-learning programs allow users to create images based on prompts and generate them relatively fast – usually within seconds.
Below are some images generated by Stable Diffusion:
Implications & Consequences:
These images are jaw-droppingly amazing but terrifying – how can humanity outdraw AI in the future? Many are questioning the ethics and legality of the image generator. The social, cultural, and economic impact that this AI could have is unfathomable. Since the code is open source, many of the code’s restrictions can be bypassed. This could have significant repercussions, like deep fakes of celebrities and political figures being made and distributed. It will be difficult to tell the difference between imitation and reality.
The entire artist community also fears for their jobs, as the “AI image generator appears to be capable of making art that looks 100% human made” (@arvalis). Twitter user @arvalis also posted that he was “extremely concerned” about the current state of art creation online. Many of the images also seem to take inspiration from art created in the past, leading to the questioning of potential copyright issues.
The Future:
These images are already very impressive and crazy to even think about. And since technology develops exponentially, it’s crazy to imagine how far AI image generation can get in the next few years. The future is uncertain; we can only see where AI and machine learning will go.
Sources:
https://stability.ai/blog/stable-diffusion-public-release
https://twitter.com/arvalis/status/1558623545374023680