How AI is shaping Art: prompt experts will be the artists of the future

In this post, we will explore AI-generated art and its future. AI is already capable of replacing artists, so how will they adapt to this emerging technology?

Not long ago, I was introduced to a community of art developers called Midjourney AI. This community is based around an AI art generator that can create breathtaking artworks from a text called Midjourney, and has since yesterday opened its gates to the world.

Its discord server is rapidly reaching the limit of 1 million users, demonstrating the huge interest that this new technology is having on the worldwide art community.

In this article, I will not focus on explaining how AI can generate art, but I will rather analyze the future of the industry.

Let’s give it a try

Before explaining how the market around AI art generation is structured, let us look at a practical example of how AI is generated. To activate the model, you will need to send a “prompt” which is a text that the AI will interpret to generate art:

prompt: an alien landscape, craters, fog, trees, artstation trending, cloudy, –ar 2:1 –w 3000
Image generated by Kenneth Scott, retrieved from this link

Other beautiful examples can be found online, all generated using Midjourney AI:

Image generated by Calder Moore: retrieved from this link
partial prompt: McDonald’s in Underwater
Image generated by Jeff Han: retrieved from this link

As you can see, this leading technology can fully replace digital artists with just some extra work (it may require additional editing with image editing software like Photoshop). There is no point in denying it, in the next three years, the industry will be dominated by AI-generated art.

What are the leading AI models that can generate art

The industry is now dominated by two leading models that compete on the same ground:

  • Midjourney AI
  • Dall-E

Midjourney AI, unfortunately, has not released any detail about the way its build, not even an architecture.

OpenAI, on the other hand, shares more details: it is based on a paper published under the Cornell University, and has been developed and will soon by made commercially available by OpenAI.

Both because of the revolutionary implications of these models and their immensely high training cost, neither of these models are open-source. In reality, compared with all the other AIaaS (AI as a service, meaning that they cost for the amount you are using them) ever released on the market they are quite expensive.

MidjourneyAI pricing

Dalle-2, has not yet released any details on its pricing, but considering the OpenAI pricing so far, it is fair to assume that it won’t be cheap.

The value in prompts: each attempt is expensive

Testing any other kind of AI, like NLP models able to classify or even generate text, is usually free, because of the minimal cost of a single processed line of text. For example, to run NLP models, a GPU is not required (although it always helps). Image generation, on the other hand, requires a gigantic amount of computing power that only one powerful GPU can provide. Image generation enters the domain of Computer Vision, a field of AI that deals with much harder tasks compared with text and number processing, raising the barriers of entry for whoever wishes to enter it.

Midjourney AI, for example, will require minutes to elaborate a single prompt and generate art, and most prompts will not generate the desired output. Especially if you are new to experimenting with the software, and generating a proper image you will need a good amount of trial and error, for good results even 10, maybe 20, or more prompts. AI is a black box (especially huge models), and we have few ways of controlling it. Each prompt is costly, so obtaining a good results while minimizing the number of attempts is of extreme value for the emerging industry.

MidjourneyAI discord server

However, the most successful “ai trainers” have experimented with thousands of prompts already, and have a good idea of how to control the software, obtaining the best output from the software. The industry has just begun, but in one year already we will likely see that the artists who are able to generate the best artists are the ones with the ability to “communicate” with these software.

A new job on the rise: the prompt expert

Following this reasoning, while the industry keeps expanding there will be a need of experts and consultants in the industry able to play with models, help companies generate art, and guide millions of followers with their suggestions.

Image generated by Elsa Bleda: retrieved from this link

The barriers of entry to art, for the first time in human history, are being leveled down, meaning that the world will have access to art without requiring trained artists: now, everyone can be one… as long as you can write prompts!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: