It’s been too long since I had a good rant about the limitations and associated foolishnesses of AI, and by AI I mean the generally available stuff. Generative language models. That which takes any and all available information (with and without the copyright holders’ permission) and answers requests using this information.
It’s tempting to think that with all the world’s information available, there is nothing AI couldn’t answer. But can it come up with creative ideas? The answer hinges on what a creative idea is. By creative ideas, I don’t mean ‘draw a cat in a meadow’ but rather drawing a cat or a meadow that startles and informs people about cats, meadows, or life, the universe and everything. It’s new. An unconventional approach to putting existing things together.
Already, I’m sceptical that AI can do this. The way the generative language models work is to put together information in ways it finds already exist. The opposite of creative.
Point 1: The basic way AI functions is not creative.
Creative people base their creativity on their personal experiences. In entrepreneurship, it’s accepted that the majority of opportunity spotting comes from the entrepreneur’s own experience of a need or gap in the market. We appreciate visual art, music, and literature because it is told with the author’s unique perspective and insights.
Point 2: AI does not experience life, or have a unique perspective. If it claims it does, it’s stealing it from somewhere.
We look to creatives to take us out of the mainstream, to recognize and communicate truths that everyone is experiencing but no one has described or articulated. AI cannot find truths that don’t exist. It cannot find gaps in our knowledge because they aren’t in our knowledge.
When AI finds gaps in the database it has, it makes shit up. It fills in gaps with nonsense. This is about as anti-creative as it gets. Rather than acknowledge there is a gap, AI inserts the intellectual equivalent of styrofoam packing chips, which were banned years ago because they are essentially evil.
I suppose making shit up might be considered creative, but that does a disservice to creatives, who do make stuff up, but in a thoughtful and meaningful way. Creatives tell us truth as they make things up.
Point 3: AI tells us misleading shit, rather than revealing truths.
AI can be told to put things together in a non-conventional way, like ‘draw a cat in an usual place’ and it might put a cat in the desert wearing a purple scarf because of the zillions of cat photos on the Internet, relatively fewer show a cat on a sand dune or wearing scarfs.
Point 4: AI will not know if what it creates contradicts the fundamentals of cat nature, fashion, or the relationship between cats and humans. The human creative might depict an unusual place for the cat as the emotional struggle a person has giving up their cat because they become unhoused.
In various ways, creatives1 will describe the process they use as an altered state of consciousness, a state where inhibitions or the bounds of tradition are abandoned to give birth to the next masterpiece of revelation. Many talk about a zone they get to where there is no other reality than the creative place (story, image, song). I’ve been there, writing fiction and prose, so immersed in the writing that Nothing Else Matters 🙂 .
Then, the discipline of the craft, whether it’s the laws of grammar and editing, visual composition or story telling or theories of business strategy or entrepreneurship, are applied to hone the concept into a consumable masterpiece of wisdom, entertainment or functionality.
Point 5: AI does not go to a zone of creativity and ignore the world, emerging with something wonderful. It might be able to apply the rules of writing, drawing or capitalism to polish a concept, but it lives among us, not in its own world.
Creativity is inextricably linked to emotion, because, how could creativity be cold-blooded, calculated and devoid of feeling? The wild careening of emotion (because emotion always seems to be wildly careening) is a fundamental part of creativity. Describing what a refrigerator, or bland apartment building, looks like doesn’t require an emotional response. AI is quite capable of doing this, although it might occasionally embellish and put seashells on the face of the fridge, but this would be random, rather than insightful.
Creative emotional content found in creations: Angry. Sad. Confused. Disappointed. Frustrated. Horny (not sure this is an emotion, but certainly inspired a lot of rock songs.). AI is not using these as motivating forces to create images, songs or novels.
Point 6: AI’s creative motive is to do what it’s told to. Now how creative is that?
And while on the subject of emotion, another difference between AI and human creatives is that humans care. AI is a machine and doesn’t care, even if it can spew words that sound like it does. From my experience, creatives want to create, abandoning time that could be spent doing something practical that earns them money, to make works that please, provoke and amuse their audiences. They do this because they care about delivering important messages to people, to inform, comfort, or discomfort for good reason. To promote societal advancement.
Point 7: AI doesn’t care what impact it has on its viewers. It answers questions or completes tasks assigned.
As eloquently reminded by Noah Yuval2, humans create abstract concepts, like money, corporations, supreme beings, and NFTs. Can AI abstract? The answer (as I see it) lies in whether abstraction is just putting existing things together in new ways or coming up with purely new concepts. This is the core of creativity. Of the zillions of possible combinations of existing things, which AI is capable of browsing through, are there abstractions that aren’t there, that aren’t logically, or even illogically? Abstraction is looking at reality and finding new concepts to pull it all together and make it easier to understand. AI might recognize trends or similarities but can’t extrapolate beyond reality. Don’t think AI could come up with explaining thunder by ascribing it to God bowling (which is where I believed thunder came from when I was a child3).
Point 8: AI can’t abstract.
As a human who is capable of abstraction, it permeates my creative works seamlessly. I use metaphors relating unrelated things in interesting ways. I anthropomorphize, extrapolating the experience of humans to understand others. I participate in human abstractions, like political parties, the influence of influencers, ethics and cryptocurrency.
For now, AI is less capable than humans at creativity, because it only reuses information, doesn’t care, or have a stake in what it does, has no goals and can’t abstract. It’s sure to evolve and become better at some things but, will it ever be as creative as humans. I’d say no, or I’ll eat this blog post. Can AI to do that?
1 I’m basing this on my own creative process, as well as living with and socializing with many creatives, and listening to others talk about their process.
2 I read his second volume first, but now am enjoying Sapiens, marvelling at the seamless integration of a whole bunch of disciplines, such as genetics, history, anthropology, zoology, linguistics, archeology, sociology and many subdisciplines I’m not properly reflecting. All good science, rolled into a single story, relatable to life.
3 Interestingly, now that I am an adult, I know that as a species, we still don’t entirely understand the physical basis of thunder and lightening.