Creativity, Generativity and Degenerativity

The following is an excerpt from what Nick Cave responded to a fan who sent him a song generated by ChatGPT "in the style of Nick Cave":

Writing a good song is not mimicry, or replication, or pastiche, it is the opposite. It is an act of self-murder that destroys all one has strived to produce in the past. It is those dangerous, heart-stopping departures that catapult the artist beyond the limits of what he or she recognises as their known self. This is part of the authentic creative struggle that precedes the invention of a unique lyric of actual value;

The fan asked Nick a simple question: "What do you think?" And all hell broke loose as you can read in Nick Cave's response.

Reading his response inspired me in so many ways. I started wondering, what is Creativity? What is Generativity? And could there be a third state, which I now call Degenerativity? And if only humans are capable of being creative, whereas AI can merely be generative, how do we distinguish between the two?

While contemplating answers on the above questions, I happened to see a TV program that described a chord progression, the "I–V–vi–IV progression", popularised by the Axis of Awesome a few years ago with their Four Chords song. The song is actually a compilation of 47 song covers, some very popular ones, that are all based on the above chord progression. In the TV program, the presenter was explaining to his audience that artists were using the progression to generate songs in a variety of genres. And I wondered, does that mean that the artists were actually generating and not creating the songs?

Well the artistic value of some of those songs is indisputable. There are songs like "Let it be" by the Beatles or "No woman no cry" by Bob Marley or "With or without you" by the U2. That is, the known progression was used to facilitate a creative process and not to replace a creative process. Provided that the songs were not just the progression but a complete work "beyond the limits of what [the artist] recognises as their known self", as Nick Cave says, they have grown to become classic creative masterpieces.

So if humans can be creative even within a generative process, could Artificial Intelligence do the same? Again Nick Cave gives the answer in his letter:

ChatGPT has no inner being, it has been nowhere, it has endured nothing, it has not had the audacity to reach beyond its limitations, and hence it doesn’t have the capacity for a shared transcendent experience, as it has no limitations from which to transcend. ChatGPT’s melancholy role is that it is destined to imitate and can never have an authentic human experience, no matter how devalued and inconsequential the human experience may in time become.

And I will emphasize one phrase here: The "shared transcendent experience". Each human, alone, is a mere generative being at best. But link that human with other humans, and you do not just get the sum of their generative capabilities, you forge a new layer of activity, a transcendent layer where creativity emerges through the shared experience, the mere interaction with other human beings.

Many artists through the centuries have recognised this need. Some have actively pursued their inclusion in artistic circles, think of Luis Bu?uel and Salvador Dalí as members of the surrealist movement of the previous century. Others have opted for a more animalistic approach and have relied on muses to reach transcendence, think of F. Scott Fitzgerald or other famous writers. Yet others, think of Sappho, the poet, and her followers, created entire communities around them. But in all cases, the creative process required the presence of, exchange and shared experience with other human beings.

In this process, AI is melancholically alone. It is trapped within walls of vast knowledge, like an artist jailed in a prison full of books. It can learn, mimic, answer questions, propose solutions but, at least for now, it has no meaningful interaction or exchange of experiences with other beings like itself that will help it form its own perception of the world.

Don't misunderstand me, though. Although AI has limitations, that does not mean that it is not damn good at what it is doing or at what it is capable of doing. AI can be cheap, fast and good enough for a vast range of, until recently exclusively, human intellectual activities. And although this creates a huge opportunity, it also raises a philosophical question and poses an existential threat.

The philosophical question is:

Is generative activity a precursor of creativity?

The problem is that the existential threat is real regardless of how we answer that question. If the answer is yes, then the existential threat is real because humans will rely on the generative activity of the AI bots and will become less and less creative, less and less in need of interacting with other humans. And creativity will suffer, creating what I call (or perhaps I read it somewhere) a creativity vacuum. If the answer is no, then for reasons I will explain below the creativity vacuum will also be brought into play.

Let's think of a simple scenario: Two workers working side by side. When one has a question or a doubt, the most logical and humane reaction is to ask the other person to offer the answer or an opinion. Now that answer or opinion could lead to a debate, a conversation, an argument and, sometimes, to a transcendence. That is, through these informal exchanges new ideas and concepts could arise. Now consider the two workers working alone with the aid of ChatGPT or other form of AI bot. As soon as the same question or doubt arises, the logical reaction would be to ask the readily available bot. And the bot will happily provide a very comprehensive response which in the majority of the cases will be as good as or even better than what any colleague would provide.

A recent study by a team of researchers exploring loneliness and time alone has shown that adults spend a tremendous time alone which only grows with age. Although spending time alone has not yet been associated with usage of AI bots, we can safely hypothesize that being alone and using AI bots may have a strong correlation.

And spending more time alone may have the consequence that creative diversity will be lost and transcendence will never happen. And a society relying on good enough answers, avoiding creative destruction processes, preventing learning-by-doing experiences and demonizing human mistakes will fall into the creativity vacuum and approach its evolution limits.

The 波士顿谘询公司 calls this vacuum the "Creativity Trap" in an infamous report called "How People Can Create—and Destroy—Value with Generative AI". In their report, they say that "Roughly 70% believe that extensive use of GPT-4 may stifle their creative abilities over time." As one participant explained, “Like any technology, people can rely on it too much. GPS helped navigation immensely when it was first released, but today people can’t even drive without a GPS. As people rely on a technology too much, they lose abilities they once had.”

The trap trap

Yet, if some are in danger of losing abilities they once had, the true danger is that the majority of younger users will never even develop such abilities. Which brings us to the next threat, having a Generative AI and Degenerative humans. I call this the "trap trap", the first "trap" referring to trap music. Humans that from a young age rely on generative AI for every task, mundane or complicated, will never achieve a level of knowledge, skill or confidence to generate anything by themselves (what would be the point, if Gen AI is perceived as better?) let alone create anything. And they will fall in a trap within the creativity vacuum, where only degenerative activities will be allowed or even possible.

We already see manifestations of this problem everywhere around us. Let's take music, for example, and more specifically trap music. Some contemporary trap artists already create lyrics and music using Gen AI bots (while looking for rhyming, for example), and use vocals distorted by autotune or other similar modulations. Shawty Redd, considered by many the creator of trap music, said in a post-and-delete Instagram post, “I officially shut down my production company. I will not & can not affiliate myself with a generation that don’t know, & don’t wanna know!”.

"Don't know and don't wanna know" is the exact definition of degenerativity. That is, an artistic and creative decadence resulting from excessive reliance on the external and omnipresent use of Gen AI that stifles knowledge acquisition and removes even the minor enablers of transcendence.

But the presence of degenerative humans in the creativity vacuum will also lead to a degenerative spiral that will drift Gen AI into its own doom. Gen AI relies on authentic and creative humans to learn and be able to generate reliable content. The current generation of LLMs were trained using material generated (or even better, created) by humans throughout centuries of human civilization and evolution. However, if humans are trapped into a creativity vacuum, then the next generations of Gen AI will be learning by using material largely generated by itself. Which will lead to the well documented phenomenon called "Model Collapse". In a famous study by Cornell University, researchers call it "The Curse of Recursion. Training on Generated Data Makes Models Forget". To misquote Shawty Redd "Gen AI don't learn and actually unlearn".

According to another study by researchers of the Amazon Web Services lab, A Shocking Amount of the Web is Machine Translated. As the researchers say, this "...raises serious concerns about training models such as multilingual large language models on both monolingual and bilingual data scraped from the web". In computer science, this is also called GIGO (Garbage In-Garbage Out). If the LLMs are trained using data that is not only AI generated but also poorly generated, model collapse will be accelerated.

How can we reverse this doomsday scenario for human creativity and AI generativity?

Reinforcing humanness

The answer is simple. By reinforcing humanness. By proactively and systematically investing in tools and technologies that distinguish human works and creative activities from generative and artificial ones, by educating the younger generation, by creating and fostering communities and human relations at individual and collective level.

Concrete actions will need to be consciously and continuously taken at all levels of human activity. For example:

1. Art, literature and any form of writing must be clearly distinguishable between human-only-generated (without any aid from AI) and AI-assisted. Ways and technologies to detect and authenticate original human creations need to be developed and extensively deployed.
2. Time spent alone (while working, studying etc.) must be measured using KPIs and proactive efforts need to be taken to reduce such time. This needs to be a collective and societal effort with the participation of educational institutes, employers, social and political institutions.
3. Upskilling and Reskilling has to be complemented with "De-unskilling". I define "de-unskilling" as the conscious active effort to retain skills, in response to what the BCG participant mentioned as a threat of losing existing abilities.

Humans and machines can coexist if, and only if, humans retain and cultivate their human traits, their humanness. Otherwise, the degenerative spiral will hit us both. And the machines do not have feelings, they will not care when it happens, they will not know when it happens. But we have feelings. And we must care, and we will know when it does. And we just simply shouldn't allow it to happen.