I love this clip because it highlights just exactly where we are in the AI revolution. Rewinding back to 1993 and how the internet looked then, you can only imagine where it will go from here with AI.

The Technological Singularity has been a popular concept in the field of AI.

Definition - The singularity is a hypothetical future point in time at which technological growth becomes uncontrollable and irreversible, resulting in unforeseeable consequences for human civilisation.

Certain technologies evoke this. When we discovered fire, for example, there was no turning back. Was it’s growth uncontrollable, sure. Once you have a paradigm shifting alteration in how we live as humans, exponential growth follows.

In 2023 we have an impressive suite of AI models that seemingly can do anything when it comes to our Digital Productive & Knowledge Base World.

William Blakes Poem excerpt:

‘To see a World in a Grain of Sand
And a Heaven in a Wild Flower,
Hold Infinity in the palm of your hand
And Eternity in an hour’

Illustrates a fundamental part of how the universe appears:

There are vast, boundless concepts that can be found in the smallest, most mundane aspects of life. Every little aspect contains the infinite.

You only have to crouch down somewhere in nature or even the side of the road and peer into the weeds, grass and various ground cover plants to see an infinite array of life. Not to mention the tiny bugs and microbes further down the scale.

How does this relate to where we are at with AI?

We now have the technology to create whole new layers of applications to everything we do in the digital realm. There are enough models to tweak, refine and purpose build to fulfil all our needs. These will do all the mundane tasks (not human replacements), help us do the more complex ones better & faster and some will do tasks on our devices not possible before like monitor, prompt & provide insights into our health.

We have seen the two main forms of models available that are Language based (LLMs) and Image based (Diffusion) show their superpowers.

Many applications can and are being developed that boot strap off of these current AI models, some big and some small.

Large ‘God-Like’ Models - Mostly Closed Source from companies like Google & OpenAI focusing on AGI models comprised of several smaller ones. These will be all purpose and have generalised capabilities. This would be Gemini & ChatGPT and Midjourney & DALLE. These utilise the full capabilities of the raw models with large datasets and even combine to be multi-modal.

Small ‘Little Angel’ Models Fine Tuned - Mostly Open Source and focused on very specific tasks and use cases. Since open source is available to everyone there will be an army of product developers working to fill every human need that can be enhanced or in the case of mundane tasks superseded by AI. These utilise smaller models that are fine tuned with smaller datasets.

I believe the ‘Little Angel’ ones will have the greatest impact in the next year and will form a kind of explosion of AI integration. These will be the vast, boundless concepts that can be found in the smallest, most mundane aspects of life.

You can look at the singularity as AI infiltrating everything in the digital space, evolving, reshaping and expanding the capabilities of what we do there. I believe this will be a very exciting time with no one really knowing exactly how it will all unfold. One thing is for sure it won’t be boring.

Magnific.AI is an example of an application leveraging an existing model for a specific use case rather than using the model’s full raw capability. In this case Magnific uses a Diffusion Model. Diffusion Models are responsible for creating amazing images from a text prompt. See Midjourney, Stable Diffusion and DALLE.

What Magnific does is fine tune it for the specific purpose of upscaling and adding detail to otherwise low res files, rather than using the diffusion model for complete ‘imaginative’ image generation.

Magnific resolves a few key problems:

• Upsizing AI generated imagery that is usually only ~1000px

• Adds more detail to any artwork based on various settings

• Generally adds sharpness and detail to any type of image, illustration or photo.

It’s a great use case and has already proven popular.

This kind of tech is open to anyone with an idea and a team with the knowledge to create great Datasets & Fine Tune AI models.

In short, Datasets are a collection of the best examples that relate to it’s final purpose and Fine Tuning is making small adjustments to achieve the desired outputs. There are various techniques and protocols that can be utilised in this process.

This means there are many possible use cases for models to be wrapped for consumers and enterprises. I predict 2024 to be a very crowded year for such offerings.

Although you can simply sharpen images should you wish, there is range of settings you can set in Magnific.AI including creative ones that add some image generation. Many users have created striking & fun examples of where it can go with the various settings.

Imagine creating a whole 3D model from a few photos.

Here’s a great video by Bilawal Sidhu showing how far this tech has come this year.

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.