What we have this week:

• Hit me with the Latest 👊🏻

• Have new Diffusion Models Taken Human Creativity’s Aura?: An Intro

• Prompt Like an Artist not an Amateur!

• Infinite Tomorrow’s Apes

• The future is in ‘Illions!”

Latest AI News

1. Apple have revealed how much they are spending on AI and believe their advanced language model Ajax is better than Chat GPT.

2. Time Magazine have recently focused on AI with their 100 influential figures.

3. Also, in an article by Mustafa Suleyman Co-Founder Of Deep Mind and now Co-Founder & CEO of Inflection AI talks about the impact of AI on society in a prelude to his new book ‘The New Wave"‘.

4. A group of Artists have signed an open letter to Congress saying generative AI is okay, arguing that AI tools lower barriers to creating art and that artists themselves should be included in the regulatory process.

5. Researchers have put forward MVDream a new multi-view diffusion model that that can create 3D models from text. It’s resolution is limited at this stage just like 2D ones were in the beginning, so this latest research from the team at Bytedance offers some great potential.

The Imitation Game Part 1 :🙈 👀 :

So what is happening?

The magic of prompting Diffusion Models with a few words and getting back a highly resolved professional looking piece of art has been an astounding revelation.

Question: Are Diffusion Models Artists?

When you put it like that it’s pretty obvious, right? The answer is a resounding no!

Let’s look at the process of art creation to draw some conclusions as to the current state of visual creativity of humans and AI.

What are the key ingredients of creative intelligence you need to produce visual artwork?

Just for fun let’s try giving Diffusion Models a Score card against Humans!

Disclaimer: It’s basically a dominance factor out of 10 so it’s not 100% accurate, but it gives an illustration of the current gains in the tech.

Diffusion Model v Humans Emoji Score Card!

Idea 🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨

Imagination 🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🤖

Memory 🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🤖

Perspective 🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨

Conceptualisation 🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨

Language 🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🤖🤖

Interpretation 🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🤖🤖

Invention 🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨

Image Recog 🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🤖🤖

Pattern Recog 🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🤖🤖

Execution 🤖🤖🤖🤖🤖🤖🤖🤖🤖😳 

Fingerprint 🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨

Expression 🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🤖🤖🤖🤖🤖

Authenticity 🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨🧑🏽‍🎨

It’s clear from the score card that visual art creation needs humans. In fact keeping a score card is perhaps not the right way to look at it.

It doesn’t have to be a competition believe it or not!

AI is making in roads into parts of the creative process, but it’s not the whole picture.

Executing at light speed with infinite iterations!

This is where the perceived magic lies.

The skills required to do the physical work of execution is time consuming and requires years of skill acquisition, training, practice and technique.

AI is now providing a shortcut and it’s a huge one!

The speed at which Diffusion Models can execute an image based on text and iterate variations of that form is the real superpower here. It’s like a lightening rod to realising your ideas.

It’s also a great opportunity for those without the forming skills to see their ideas realised. A little democratisation of art.

Next week we will continue with some further insights into Human and AI creativity and talk about it’s super powers, it’s limitations, state of IP and give some tips on how to utilise this potentially game changing technology!

Prompt Strategy for Artists

Here are some approaches to working with Diffusion Models creatively to better realise your artistic vision.

1. Develop your Idea First

Put everything you can into developing an idea. Do the work! Don’t just go with the first thought you have. Think about your inspiration and explore your idea’s ‘world‘ further. You can even use a Diffusion Model as your sketchbook.

The stronger your idea to be begin with, the better your work will be!

2. Be Specific

Diffusion models are labellers. They will take each word and in some ways mysteriously reconstruct from the combination of labels you give it. Give each specific thing you want to see and try different synonyms of that word to gain better results.

3. Iterate, Iterate, Iterate

Think of the process of image generation as just that, a process. The more you iterate the better and hopefully more original the result. If you are going off the first prompt you are probably getting an obvious imitation at first.

4. Game the System

Once you have something in mind. Look at doing a staged approach to your prompts and this will steer the model to your ultimate genius vision!

Don’t always start with the things you want to see in the first prompt. You could start with shapes and simple forms to define the composition. With the diffusion model’s regenerating each time from the previous you will have more control and creative collaboration to get a great result.

Below is an example with possible [sub]prompts. You can have more in each prompt stage, but this gives you an idea of the through-flow process.

There are infinite ways you can tailor a process like this.

5. Combine and Conquer!

A Diffusion Model will do its best to represent your text and will combine things in all sorts of wondrous ways. This is an element of art that is powerful in conveying ideas.

Try and be as original in your combinations as possible. Think outside the box! Also, try combining a couple vs many.

Banana + Day 🍌⏱

Banana + Nurture + Crane + Day 🍌👩🏻‍🍼🏗⏱🤪

Think of your idea and all it’s complimentary and opposing elements.

6. Collaborate

Try adding your own self made images. This is a way to collaborate and influence the output with some of your own original material and put your fingerprint on it.

7. Use More Than One Diffusion Model

Now that you have found a great new tool, why limit it to one! Each model has been fine tuned in different ways, so be sure to run your ideas through as many as you can and even upload a result from one into another.

8. Experiment

Try poetic, abstract and downright crazy illegible gibberish if you like. See how the model interprets and then do the above. It’s a great way to free constraints and find unexpected paths to express your ideas.

This will also ultimately help you understand the model and it’s behaviour. The more you do, the more command you will have over your results.

Infinite Tomorrow’s Apes

Been busy creating this series of Tomorrow’s Ape’s with ideogram.ai

With AI you’re able to make endless iterations with each Ape seemingly having their own individual personality!

It looks to me like the future is going to be in ‘illions’!

What did you think about this weeks Newsletter?

Did we nail it? Can we do better? Want to know more? Hit reply and let us know!

If you like our content and think it might be worth a shout out, please do!

Right click the below button and share the link.