Brand protection company and domain name registrar, MarkMontior registered at least 79 domain names yesterday that are what some domainers call brandable domains; 5, 6 and 7 letters that do not spell a word.
They do not seem to be typos of famous brands and almost all were hand registered.
All of the domain names we tracked are .com’s
Here is the list:
Horrible names wow, I hand registered Rigza and Micza. There’s better “Brandable” domains you can register right now…
Asus + io
Asus + iq
These are AI-generated names using a plurality of criteria, including but not limited to length, parsing, word-focus (vu, qu, op, be, ke, xu) and availability.
More exercise, than creating possible brands or names.
Client could be Google, or pharma, or who knows?
The big Branding companies will be replaced by AI. soon. Rather than paying the Brand Institute, Landor, or any of these companies $500 K to come up with a set of possible names for a product or service, AI will create the names, then use AI-generated focus groups.
Maybe even Mark Monitor will add branding to its offerings, and this was a mere test. All horrible names, so there’s still work to do.
Joseph Peterson says
A large part of what I do professionally is program that AI (where domains are concerned). The human element in naming will never be replaced except with bad results. Naming is fundamentally an act of creative writing on a very small scale. Even the act of choosing between names is an act of intelligent reading. And that’s something machines do very very badly – especially with short text, which may be densely symbolic or allusive.
None of us foresees a day when novelists, poets, movie screenwriters, and standup comedians will be replaced by robots. Computers excel at the task of generating name ideas or even prioritizing options; but they’ll always lag behind humans when it comes to assessing meaning, making aesthetic judgments, or having a sense of humor. Knowing when to laugh and when to groan requires a HUGE amount of socialization.
No robot can read “The Waste Land” and make sense of T. S. Eliot – let alone judge the writing as good or bad. There’s too little material in the poem itself. The context lies largely outside the text. AI would need to mine all the literary criticism ever written in order to make a guess about what the author is driving at. And that’s not even a question of quality, which is far more difficult to gauge.
The content of a brand name is boiled down to something impossibly small – just a word or 2, or even a single neologism. Computers CAN make inferences about such small inputs. But that requires massive resources and computing time to do at even BADLY for a single name idea. To come up with crude assessments for all the names the computer so easily generates – that would be a process running a million or a billion or a trillion times longer.
Nobody rents time on supercomputers to tackle problems like this, since human beings are readily available to respond to brand name ideas; and we’re far cheaper than supercomputers, which are busy modeling chaotic systems like the weather and climate change.
M. Menius says
@Joseph – “Computers excel at the task of generating name ideas or even prioritizing options; but they’ll always lag behind humans when it comes to assessing meaning, making aesthetic judgments …”
Excellent point, and relevant to the complexity of domain appraisal. The bot-driven appraisals sold at the registrars are of little use (estibot included) because there are complex factors in the human assessment of quality that machines can’t duplicate.
The key question is of course: will Mark Monitor submit these to BrandBucket?
Voice As Computing says
Great post by Joseph Peterson!
Internet technology comes in leaps and bounds, Voice As Computing being one of many, but humans are here to stay: to create and consume.
Mark Monitor may have reasons for those hand-reg,but i think there still lots of better 5L names
available for reg that are brandable. e.g , i got OMIZY, AZIPI,ONOFA and more.
Mark Monitor should search deeper for better brandable names in 5L
Short Brandable Domains says
I always keep reminding myself that some acquisitions might be completely long term, maybe they expect a return in 5-10 years. Other than that, my guess is as good as anyone’s. Perhaps we can learn more by looking into similar patterns (they registered obecva.com and obecvu.com , but obecvi.com is available. WHY?).
I used my tool to register a few similar ones, just for fun. Like scampo.com, oebar.com, corsap.com, rihor.com… Right now, I think there’s a ton of domains like that.
I agree with you on some points.
But I’m working with IBM Watson and Google’s DeepMind.
AI-journalistbots already create a large percentage of sports, financial and “breaking news” reports.
An AI bot has created several poems that doctorate students at Cambridge, Oxford & Princeton thought were written by either Wallace Stevens, T.S. Elliot, or Elizabeth Bishop
AI-Bots are creating jazz music that critics have compared to John Coltrane, Miles Davis and Duke Ellington
Of course algorithms cannot replace that “human spark”, and least not now.
As far as branding, these companies – Google, Facebook, Amazon, Apple, IBM, Novartis, Microsoft, Dell, Samsung, Tesla — are already experimenting with AI-generated name creation. This will be harder to achieve, than say, mashing Eliot, Stevens, Lowell, Yeats and adding a plurality of other criteria to create a faux “Wasteland” or “Sailing to Byzantium”
Joseph Peterson says
Developing AI to pass the Turing test (i.e. to pass itself off as human) is little more than an academic stunt – and a stunt, furthermore, that requires massive resources and a large amount of human labor. AI is only useful if it tackles problems BEYOND what a human being can easily do – problems like climate modeling, say. Producing bad jazz with a synthesizer or gobbledygook free verse to confuse readers is a waste of resources.
Using AI to generate name ideas is easy enough a 12-year-old with a “for loop” can do it. Judging name quality and suitability is much harder. Even that’s a soluble problem, given enough data and computational resources. But my point is that this problem isn’t actually a problem at all. Humans can discriminate between good and bad brand names far more efficiently and accurately than a computer. And there is insufficient financial motive to develop the technology anyway. The domain market itself is too meager a prize to divert supercomputers from, say, predicting stock market behavior.
Do you have links for those stories? Graduate students who don’t know their literature and can’t tell the difference between poems by T.S. Eliot and Elizabeth Bishop might well fall for cryptic computer-generated verbiage – that just mens they haven’t read the authors in question or only pretend to understand what they read.
I’m skeptical and wouldn’t be very impressed by that sort of thing even if it were true. I mean, a music critic might say, “Yeah, the elevator muzak I’m listening to sounds a bit like Miles Davis / John Coltrane” because the programmers designed their AI to mimic “Kind of Blue” or “Giant Steps”. After all, they’ll have to say it sounds like SOMETHING. And it will be pretty obvious what major landmark figures the AI has been trained to imitate.
From your lips or taps to God’s ears.
If we start Ai-sourcing “creativity”, I may be heading towards the bread lines.
One of my colleagues created this:
Easy on the ears, but no “Kind of Blue”…
Joseph Peterson says
Listening to it now. Genuinely it’s pretty cool as a novelty experiment. I heard 1 track that was unpalatably syrupy … with a metronomic drum machine doing it no favors. But another track (solo synth piano) actually holds real musical interest.
The reason, I think, is pretty telling. That’s because the melodic contours, phrasing, and (lack of) coherent thematic development are so far OFF from what a typical human musician would do that they become rather intriguing. Actually, a lot of jazz musicians strive for angularity and rhythmic dislocation, avoiding what’s predictable. So that track succeeds BECAUSE of its failure to imitate what’s idiomatic. Then again, I’m a fan of Cecil Taylor! (Used to own CecilTaylor.com, in fact.)
Cecil Taylor, while being far more dissonant, is recognizably human and ultimately more interesting musically because he pursues a human idea, developing it in a way that appeals to humans.
This AI-generated music isn’t so inventive, really. It begins with a Pat Metheny track (or multiple tracks) of actual music. Then it manipulates it to produce something slightly different. To me, that seems rather like performing a large number of Find-and-Replace operations. Structuring a new work is the real challenge. A computer can rewrite “War and Peace” replacing nouns with nouns, verbs with verbs, adjectives with adjectives; and it might ultimately look almost real. But this disrupts lines of character development, the flow of narration, conversations, etc. The result is gibberish, even if it retains some of the “form” of the original. And it isn’t really built anew. It’s just filtering an input to get a slightly different and inferior output.
The Metheny-derived tracks don’t make sense (in which case, they’re interesting) or they sound insipidly “normalized”. At neither extreme do they recreate the kind of human interplay we’d hear between Pat Metheny, Bob Moses, and Jaco Pastorius on an album like “Bright Size Life”.
Humans are out of AI’s league.