How Creative Professionals Can Benefit from Deep Learning Concepts
I’ve been thinking a lot lately about our technical advancements with AI, which has sparked many philosophical conversations (with myself).
The same motivations that drive us towards certain behaviors are also the same ones that allow the magic of Deep Learning to happen for autonomous entities.
So, I asked myself, “If by creating artificial intelligence, are we essentially trying to embed ourselves as the proverbial ghost (our consciousness) in the machine (the AI hardware)”?
I went into quite the rabbit hole.
But, what I came away with (for the time being) is, “Can we, as Creative Professionals, learn from the techniques used in Deep Learning that help AI overcome obstacles and grow?
It seems like abstract concepts of Philosophy are converging heavily with our progress in AI.
Is our “mind” or “consciousness” just a side effect of environmental conditions coupled with biology?
Should we consider AI in a similar way?
My instinct is that there is far too much at play to reduce the complex nature of systems into neat little boxes. Still, for the purpose of this article, I will explore some of the overarching concepts used to train AIs and how we can manipulate those same conditions in ourselves for our own creative growth.
Here’s a bit of a non-technical view of how Deep Learning functions:
At the heart of Deep Learning are “Neural Networks”. Just as our human brains are made up of interconnected neurons that process information through connections, artificial neural networks effectively emulate the same behavior.
To be put crudely, think of one of those telephone switchboards from the 1940s.
Artificial Neurons process information, recognize patterns and make decisions much like how our brains learn from sensory inputs (Touch, Taste, Sight, Smell, Hearing, etc.).
With Deep Learning, environments are presented where the AI can develop skills such as reasoning, detecting objects, learning how to move, and other things.
Three things really stick out to me when I watch those Robot videos from Boston Dynamics:
- They fail a lot, and it is encouraged
- Input Data is required to train them
- They are incentivized to learn
If we extrapolate the meaning of these three assumptions, we get:
- Iteration
- Information
- Incentivization
It is reasonable to suggest that the “Iterative Process” is at the top of the Deep Learning Hierarchy. Without failing, refining, and re-application, the AI will not learn.
Moving down the “food pyramid” of Deep Learning, we get to “Information.” Just as the Creative finds inspiration from the world around them, AI requires a large dataset to understand their environment. In the physical realm, that manifests in the need for many sensors to detect environmental changes.
Then, at the bottom of the pyramid, Incentivization is the foundation for the whole cycle. Again, much like the Creative, if an AI is not incentivized to complete a specific task, it won’t grow.
In my eternal search to find the line where humans end, and AI begins, I want to draw an obvious distinction between what both parties are good at. Moravec’s Paradox is an interesting read and highlights an important point about AI.
“It is comparatively easy to make computers exhibit adult-level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility.”
- Wikipedia (https://en.wikipedia.org/wiki/Moravec%27s_paradox)
To grasp the complexities AI faces in mastering human-like perception, we must acknowledge our own evolutionary history that it needs to meet.
Our intuitive tasks, whether recognizing a friend’s face in a crowd or navigating through city streets, are underpinned by millennia of evolutionary adaptations.
These capabilities are deeply rooted in the subconscious game of evolutionary “Risk vs. Reward.”
AI will exceed human capabilities in many ways, from computational speed to data processing. However, the uniquely human attributes — such as emotional intelligence, nuanced comprehension, and creative intuition — still span a bridge too far for artificial systems.
This paradox fascinates me as it vividly illustrates how our inherent traits may complement each other. However, we should also consider how we play to our own strengths and delegate tasks to people/machines who are better suited for the job.
Incentivizing Creativity
Reward Functions
Unsurprisingly, we, as Creatives, don’t make much progress unless there are some rewards. Similarly, AI has Reward Functions that encourage learning. The AI performs an action towards a particular goal. If it succeeds, it is rewarded with Positive Reinforcement, which fuels it to repeat those actions.
With Creatives, goal setting can be incredibly rewarding if achievable and measurable. By creating our own mental reward system, we can push our boundaries and maintain a certain level of sustainability to continue doing it.
Things like:
- Client Approvals
- Being recognized by peers
- Personal Satisfaction
- Money
All serve as positive reinforcement that drives us to experiment and learn on our path to “Getting Better.”
Practice Makes Perfect
Iterating and Reinforcement Learning
So incredibly familiar to the Creative, the cycles of trial, error, and improvement are vital to the Creative Process.
What I’m saying is nothing new. We all know that “Practice makes Perfect” is an adage we hear on a daily basis. But looking at it through the lens of Deep Learning, I feel we can appreciate it more.
Seeing AI’s iterative process at work is much faster than assessing our own progress internally.
Here’s a video of the Deep Learning process for context:
That old adage nicely complements how AI learns by failing at a task, gathering feedback (through automated or human intervention), and attempting to improve for future cycles.
Many Creatives are tinkerers by nature. Rather than learning theory, applying something over and over until the desired result is reached is an iterative process in itself. Just like AI being trained in a multitude of simulation cycles, the Creative undergoes the same transformation but on a much more subtle level.
As creatives, we should embrace the process by iterating on our work, refining concepts, and seeking feedback from unbiased sources. Find people who aren’t trying to make you feel better but those who will push you on your ideas and execution.
Moravec’s Paradox
Valuing Diverse Sources of Inspiration
I mentioned earlier that AI training requires large datasets. The same can be said for creativity. The great philosopher Pablo Picasso once said, “Good artists borrow, great artists steal.”
Whether AI or Humans are at the top of the conversation, this statement is a good reflection of both. Feeding AI’s insatiable appetite for data results in more robust and nuanced models, much like how artists get inspired by consuming things like culture, food, beauty, and nature. Experiencing a diverse range of things helps people gain new perspectives and broaden their body of work.
This all sustains the “Creative Drive.” Assigning value to the “Inputs” — diverse Sources of Inspiration — will allow the Creative Professional to push the boundaries of their work and find new solutions to age-old problems.
The Symbiotic Future
Integrating AI and Creative Processes
As I’ve said in previous articles and posts, AI should enhance the Creative’s workflow, not replace it. Automate away all the mundane, laborious tasks and focus on driving the concept and the creative execution.
Bingo.
If you use AI to create images or videos, make sure they look exactly how you envisioned them. Midjourney is great, but bringing that image into Photoshop for further refinement is gold.
Or do what I do and use the AI-generated image as inspiration to create the work in your style. Just because it is easier to find inspiration using AI doesn’t make it any less meaningful.
Integrating AI into Creative Processes will be an inevitable part of our future. Our charge is to maintain integrity and authenticity when using these powerful tools.
The overall vision of a project trumps any single component comprising it.
Our most powerful tools as Creatives are the emotional connections we make and our storytelling abilities, something AI cannot replace.
The best thing we can do is begin to understand how AI works. Then, we can better understand our limitations and how we may benefit from making it a crucial part of our future innovations.
The human touch will always carry intrinsic value. Let’s start exploring that now, so in the coming years, we will already have found our flow where we can stick out in the marketplace, be recognized for our accomplishments, and peacefully co-exist with AI…just in case Skynet is created.
Authors Note -
This article was written using traditional methods (me!) and grammatical corrections from Grammarly. Initial research was conducted using ChatGPT and then expanded with conventional research methods.
As a side note, I think that the ability to verify content originality will be critical soon. For the time being, I created a crude method of doing so, but I will refine it over this year and open-source the results for others to use.
Until I build a more robust solution, my digital content will be uploaded to IPFS, and a digital asset linking to it will be created automatically upon upload.
A transaction using the wallet specified below will be signed and verified on the Algorand Blockchain, which will indicate that I published this work.
Verification of Content Credential — {Link to Transaction}
Author Wallet Address — driventocreate.algo
References -
Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press.
McCormack, J., & d’Inverno, M. (2012). Computers and Creativity. Springer.
Sutton, R. S., & Barto, A. G. (2018). Reinforcement Learning: An Introduction. MIT Press.
