Generally Productive Talks #2 - Ai and Images

Introduction

Ai-generated images are on a rapid rise currently, their capabilities and input methods on a constantly increasing gradient, as mentioned in last week's posting, all levels of art and design will be impacted by Ai and it is those who learn to use it, and have the capacity to utilize it as a tool, not a final solution who will have a leg up on the competition

In today's post, I use as much as possible Chat-GPT to examine Ai generated images, their history, and their future, but also separately examine the options available today, and which are the best to use for what

The Past

The origins of Ai images began with A. Michael Noll in the 1960s who made the first Ai images, geometric patterns made with code he wrote himself

It was then the Mandelbrot set from 1980, that used the idea of creating intricate patterns based on mathematical formulas and was the early inspiration for modern Ai generated images

In the 90's General Adversarial Networks (GANs) were incorporated which use 2 neural networks, one to generate and the other to evaluate images to produce high-quality images. However, it was not until advancements in machine and deep learning technologies were made in recent years that it allowed Ai generated images to advance to the point they are today

Today

The level being achieved already is substantial and important to take note of, to understand this better I have researched and examined multiple options (that are free to use, those that charge I did not try yet) in the market today, and below give options and feedback on what you can use for what and show you how they do:

1. Art

2. Photo Realistic Generation

3. Logo Design

4. Concept Sketching

5. Concept Renders

1. Art

This seems to be the main area of use currently and all options I found supported use for art, I selected DALL-E2, Pixray, and NightCafe (Stable) and generated images based on 3 of my favourite artists, the below are the best results from various Ai generators I attempted:

Van Gogh

Starry Night style War of the Worlds setting

Botticelli

London populated by Corgi's

Jim Lee

An alien in a batman suit fighting a muscular man made of mirror material on the moon

2. Photo Realistic Generation

Using Dream by Wombo, NightCafe (Stable), and StableDiffusion I generated images of;

Marilyn Monroe visiting the Eiffel Tower (don't try to do faces for free yet...)

Xenomorph from Alien movie posing for a portrait

These were using free software, paid for software is already much better, and can produce fully generated images of fake people such as the below;

3. Logo Design

I tested many options for Logo's, most are very simple and quite bad in general, my aim was an alternative modern version of logo for NASA, I selected Looka, Brandmark.io, and DALL-E

4. Concept Sketching

The options are much more restricted for this area, no option I found was remotely useful yet but I managed to get something from DALL-E, Dream Studio, and Stable Diffusion to show their current levels of ability:

Prompt: Hoverboard designed in the style of Tesla

5. Concept Renders

Many options for this area, I selected Deep Dream Generator, Dream by Wombo, and Stable Diffusion to demonstrate the results, which I was quite happy with for what it is:

Prompt: E-bike designed in the style of the 1996 Honda Fireblade with a gold and black colour scheme

Honestly, I've seen a lot worse done by a lot of designers... personally, I would buy the DDG design tomorrow...

Even more options exist that I won't go into today, Fashion Design and Interior Design (as shown below) are possible right now and while today I only utilized written prompts, image-based design is also possible to varying levels of success

The Future

The genie is out of the bottle now and Ai's ability will only increase as we look to the future, expect that in;

5 years - continuous improvement of today's abilities, with lighting, texture, photo-realism, and the ability to blend styles to become astonishingly accurate and powerful but still requiring human direction, curation, and interaction throughout

10 years - entire technologies, programs, and areas of business will be integrated into utilizing Ai with the quality and scale of its ability to create not only images but entire videos and interactive experiences requiring critical but limited interaction with the designer, which will mean many fewer jobs in associated sectors within large organizations, while many more small operations utilizing the technology individually or within very small teams generating much larger amounts of workload than is currently possible

20 years - by this point Ai should be able to create art and designs on the level of or better than almost all humans, including abstract art and concept designs, creating entire portfolios of products, designs, and experiences, and reshaping each industry

My thoughts:

Given the above, we can see this is not going away and will change the world and design in transformative ways. These tools (although in some cases unstable) are already useful, there is clearly a lot of development still to do and a lot of trial and error is required to get something useful, but I think the timelines discussed for their advancement are conservative

The reality this presents both scares and excites me in equal measure, this is a game changer and will mean every aspect of art and design will never be the same again. Learning to use these tools and staying on top of their advancements is the new basic requirement to be able to do design and development, and finding a niche to operate with it is how we can survive

Try out those I have recommended above and try out others you find yourself too, if you are a designer or just do it for fun, you won't regret it!