People believe that being creative is a necessary component of being human. However, artificial intelligence (AI) has matured to the point where it can be creative.
Artists were outraged when an AI model named Midjourney won a recent competition. And this kind of software is now free as a result of the introduction of a comparable model called Stable Diffusion, which is the best of its kind to date.
For a long time, creative unions like Stop AI Stealing the Show have been concerned about the usage of AI in the creative industries. Can AI artists really take over?
What these new AI models can achieve is limitless. The robot pictures seen above are all unique, although they were all created by Stable Diffusion in response to similar user requests.
You may use these AI artists in two ways: to compose a brief text prompt or to add a picture to the text prompt to provide further information. Based on a 14-word challenge, it created various logo concepts for a fictitious fruit delivery service in under 20 minutes.
As the data show, Stable Diffusion has difficulty creating word-based art. Some of the fruit appears strange.
Because of how important this technology may be, Stability AI, the firm that created Stable Diffusion, is looking for a US$1 billion (£900 million) investment review. However, the award-winning Midjourney image demonstrates that AI models are beginning to have an impact on the real world. In fact, AI is quite adept at creating works of art that blend many styles and components.
While AI may handle the majority of the work for you, you must still understand how to utilize these models. Sometimes a prompt does not provide the desired result. Alternatively, the AI may be combined with other technologies as a minor component of a bigger pipeline.
And creating excellent art is not the same as creating computer designs. Stable Diffusion makes it simpler to sketch landscapes than logos.
Why Stable Diffusion is revolutionary
Most of the time, AI models learn how to make art by looking at a huge library of 5.85 billion photos. This massive quantity of data is required for the AI to learn about what is in photos and how art works. And the procedure is quite time-consuming.
The CPU needed 150,000 hours, or a little over 17 years, to complete Stable Diffusion. This may be reduced to less than a month by training on massive computer clusters in parallel (collections of powerful computers that act as a single device).
Stability AI also offers an online application called DreamStudio, which allows you to employ its AI model for as little as US$0.01 per picture. The expense of using OpenAI's DALLE 2 art model, on the other hand, would be more than 10 times higher.
Both approaches rely on the same fundamental concept: a computer software known as a "diffusion model" that learns to create new pictures by studying a large number of old photos. However, since Stable Diffusion has a lower computational cost, it requires less time and energy to train.
You also can't download OpenAI's model and run it; you can only use it through a website. Stable Diffusion, on the other hand, is a free project that anyone can use. So, the online coding community can quickly make changes like updating models, making user guides, and linking it to other tools. This has been happening since August 2022, when Stable Diffusion was released.
Our future with art
In the previous five years, AI art models have come a long way, yet they still struggle with certain aspects. The words in their work are usually identifiable, yet they make no sense. Likewise, AI has difficulty sketching human hands.
The fact that these models can only make digital art is another problem. Animals cannot use oils or pastels, but humans can. Technology, much like vinyl, may produce a swing toward a new form at first, but people always tend to circle back to the original form that was the greatest over time.
However, there is still a danger for creative individuals. Digital artists who do not employ AI may lag behind because they are unable to keep up with how rapidly AI-enhanced artists can make adjustments and how little it costs them to do so.