Creativity involves breaking out of established patterns to look at things in a different way.
-Edward de Bono
Naming is a notoriously challenging niche of branding, and, as such, holds great promise for disruption. Similar to the way human brains work, generative AI uses deep learning and neural networks to write essays, create code, draw up designs, whip up recipes, compose songs, and a whole host of other needed and needless things.
If such a tool could successfully unravel the complexities of naming, it would revolutionize the naming profession. Driven by curiosity and career concern, I, like numerous other naming specialists, tested ChatGPT and similar platforms through a series of simulated naming tasks.
The bots’ naming chops, as documented in my inaugural article on the subject, were less than impressive. And on the rare occasion when plausible names did arise, they all risked trademark infringement.
Nevertheless, given the technology’s nascent stage, it is still too early to write off the bot’s abilities. So, when Corsearch, the renowned and trusted provider of trademark research and protection tools, announced their Name Generator, I signed up for a trial, hoping the technology had improved since my last experiments. Too, the Corsearch naming generator had the added value of working in tandem with the company’s exemplary trademark screening software.
Getting over the hump
For this go around I followed the advice of the Greek thinkers – “Know Thyself” – and challenged the bot to name an AI start-up.
The Corsearch name generator, like its competitors, almost consistently produced alliterative names using the “camel-case” convention (CamelCase or camel caps) such as SynthoMind Solutions, CogniCreate Labs, and VirtuVerse AI.
I would call these VeryBad names.
So, I got more specific:
Give me 5 non-alliterative neologisms for a new AI start-up inspired by computer science history.
This command produced TuringWave, AdaLogic, BabbageMind, Lovelace Link and others of similar constructions.
Besides repeating the camel-case convention, the bot showed little imagination, basing its names almost exclusively on well-known computer luminaries like Alan Turing, Ada Lovelace, and Charles Babbage.
In an effort to break the bot’s penchant for alliteration and on-the-nose references, I prompted:
Create four non-alliterative company names for an AI start up inspired by Shakespeare’s writings on the mind.
The results? Names like SonnetSynth, IambicIntellect, FalstaffForesight, and VeronaVerse.
Despite various prompts, the bot continued painting by numbers, and insouciantly skipped over my command for non-alliterative names.
The sound of music
Of course, there’s nothing wrong with alliteration — if judiciously used, it is a tried-and-true naming device that can lead to sticky brand names. Past examples include Coca-Cola, Rolls Royce, and Minute Maid, and more recent ones, TikTok, Ted Talks, and SoulCycle.
However, the bot’s alliterative names tended to be quite literal, lacking the surprise pairings found in names like SoulCycle and Minute Maid.
Furthermore, for reasons not clear, the technology seldom uses other poetic devices like assonance, onomatopoeia, and consonance.
It also seemed unaware of how the auditory quality of words can evoke meaning. Training the bot in sound symbolism and phonology might help in this regard.
Breaking the camel’s back
My subsequent prompt aimed to increase the bot’s musical range, and break its dependence on medial caps and explicit references:
Create ten single, invented names for an AI startup inspired by Shakespeare’s thoughts on the mind. Do not use camel-case convention.
This directive broke the camel’s back and eliminated alliteration, but it resulted in tongue-torturing grotesques like Linguispeare, Prosemaestro, Sonneteerify, and Metaphrasis.
Shakespeare, the brilliant inventor of words such as “moonbeam”, “bedazzle” and “zany” (to cite a few of the purported 1,700 words he created), would be turning (turing?) in his grave.
Though in fairness to the bot, Shakespeare also created some doozies like “enacturcs” (purposes put into action) and “gallimaufry” (a complete mixture, a medley), to mention a few.
So, what’s the problem?
Many in the brand naming industry have noted that generative AI’s inability to create compelling brand names stems from its incapability to grasp the nuances of a company or product’s industry, intended audiences, mission, and purpose. But as the platform evolves, these and related issues can surely be addressed.
I am, however, more skeptical of AI’s ability to master three capacities unique to humans, and crucial to good naming.
Let’s take a look.
Off the beaten path
In the realm of creative naming, the true maestros are adept at lateral or horizontal thinking. Coined by Maltese author and psychologist, Edward de Bono, in 1967, the term encapsulates a less straightforward approach to problem-solving.
At River + Wolf, we use multiple lateral-thinking exercises to generate new ideas, including the well-known one of randomly picking a word from the dictionary. This word then serves as a springboard for creating associations that have some relevancy to the offering being named. This exercise and others of its ilk, force the mind to depart from predictable associations.
As of now, in the realm of naming, generative AI platforms tend to adhere to well-trodden paths. The bots are largely “vertical” not “lateral” thinkers. And vertical thinking seldom leads to compelling brand names.
Like many naming professionals, we research various fields in search of ideas. For instance, when working on a recent crypto project, we delved into a wide range of disciplines including mathematics, logic, astronomy, ancient games, weather terminology, deep sea diving, and the Italian Renaissance.
During this exploration, we stumbled upon the symbol (∀), commonly used in mathematics and logic to denote “for all” or “for every.”
We translated this visual symbol into language, resulting in the name, “Inverted A”.*
This symbol and its meaning complemented the client’s interest in a name conveying democracy, a valued concept in crypto and Web3 (though he eventually selected a different name).
We have yet to encounter an AI that approaches naming in the way described above.
Brand names with emotional resonance often emerge from a naming specialist’s well of memories and sensory experiences. During the naming process, these involuntary memories often arise spontaneously.
For instance, while strolling down a Parisienne side street at dusk, a snippet from the Charles Aznavour song “La Boheme” — “Accrochait des lilas jusque, Jusque sous nos fenêtres” (hanging lilacs, just under our windows) — drifted through a cafe door and dropped into my memory. Years later, this auditory experience resurfaced as “Bohemian Lilac,” a name candidate for a new paint color.
After the name arose, I recalled its source. The combinatory powers of the human mind had artfully joined the song’s title with the lilacs mentioned in the lyrics. Dusk’s lavender shadows no doubt informed the name as well.
This sensorial and experiential reservoir is something AI lacks, because in its present state it has neither sense nor emotional memory.
Considering these factors, I find it improbable that generative AI will completely replace human naming specialists. That said, if machine learning evolves from generative AI to artificial general intelligence, and is able to replicate these modes of human perception, it could give naming specialists a run for their money. (Though if AI ever reaches this stage, we may all be running for our lives.)
But until computer scientists find that holy grail, generative AI platforms are best used as naming-support tools, as highlighted in Catchword’s insightful article.
Nevertheless, even in its current stage, I believe these platforms, with some additional training, could work well in one especially challenging naming domain – pharmaceuticals.
Prescription for success
In the pharmaceutical realm, names must transcend direct communication of therapeutic benefits. Because of this constraint, “meaning” is often conveyed through sound. This is demonstrated in the novel weight loss drugs, “Ozempic” and “Wegovy.”
The former’s sound proximity to “Olympic” conjures images of vitality and fitness, while the latter cleverly combines a nod to “weight” (we) and “go,” suggesting a departure of weight. The choice of the long “e” vowel at the end creates a light, almost gleeful feeling.
If generative AI platforms could be trained to hear meaning in language fragments, and then combine these bits of language in appealing ways, it could change the way drugs are named. And if this output is integrated with global trademark databases, as Corsearch has pioneered, the impact on the industry could be even greater.
Adding POCA – a software tool that uses an advanced algorithm to identify orthographic and phonetic similarities between drug names – would further optimize the platform. Such a tool could ultimately lead to the downfall of naming agencies that focus exclusively on pharmaceutical clients.
Happily, River + Wolf’s clients span various industries. And considering the current capabilities of AI in naming, I don’t anticipate we will be closing our doors anytime soon.
* Inverted A is the intellectual property of River + Wolf brand naming agency