How to Be Different From a Robot

AI can write, yes. But can it think? Not yet

In a recent episode of Akimbo, Seth Godin posits that “he” might not be real. He/it might actually be an AI-generated robot podcast host, author, and newsletter publisher. 

I have a client whose software generates reports that aren’t “real” either. As in not written by a real human. They’re written through “natural-language” generation (NLG). His NLG packages dry data into paragraphs that look like they are written by a human. The intent isn’t to fool anyone; in fact, my client discloses to his customers the role of NLG in their reports. He just uses it because it makes the reports easier to read.

The SEO-faction of content marketing teaches us to write like this, too. We are supposed to game search engine algorithms using artificial writing and video production conventions, including:

  • Artificially favoring (but not too much) “search keywords” that connect your content to searches performed by your potential audience
  • Writing in one-sentence paragraphs to create a sense of build-up and make your content easier to scan, rather than read
  • Over-using huge letters, bold letters, italicized letters, bullet-points – all to generally avoid a series of substantive paragraphs of plain text
  • Over-using images of people’s faces, large, dramatic landscape images, and over-using images in general
  • Setting artificial constraints on length. For example: 7 minutes for YouTube videos, 25 minutes for podcasts, 2000 words for blog posts

There’s an economic logic to these rules. YouTube recognizes that on average fewer people watch a 27-minute video than a 7-minute one. Google recognizes that people are more likely to share a 2,000-word blog post since it’s difficult to create one based entirely on fluffy thinking. Difficult but not impossible, that is. One can double the length of blog post with empty phrases.

But here’s the thing:

A speaker who uses that kind of phraseology has gone some distance toward turning himself into a machine.
George Orwell

This has the effect of “plasticultura” tomato cultivation – looks like a tomato but tastes kind of like neutral cardboard.

But machines are competing hard with humans to produce cardboard content. Like 100’s of other viable, purchasable software products, Zyro CMS (similar to Wix or Squarespace) already offers AI text generation. Make a new page, title it, keyword it, and let it write your web copy the way the SEOs say to.

Or tries to.

I typed the first paragraph of this post into Zyro’s AI Content Generator as a “contextual prompt”. In response, the AI output continues some kind of narrative, at least, but not a very good one (AI-generated text in italics):

In a recent episode of Akimbo, “Seth Godin” reveals that he might not be real. He might be a robot podcast host, author, and newsletter publisher. On the show, you’ll hear from Seth’s daughter Sasha who is worried for her dad as she shares details about his unusual appearance which includes missing eyesight in one eye, two fractured vertebrae on his spine (with ligamentous damage) due to bone fractures involving bones near each other muscular weakness caused by neck/hip muscle injuries resulting when wearers were prone to falling out of chairs during their youth… 

While impressive (and also a bit hilarious) for recognizing that there was a narrative in play to build on, there are some major problems with this content:

  1. The narrative is complete nonsense. The context I gave the AI had nothing to do with medical problems
  2. The grammar is bad to the point of causing ambiguity or confusion
  3. It contains empty, vague, over-used, or unnecessary phrasing: “On the show”, “you’ll hear from”, “as she shares”, etc.
  4. There’s a lack of original writing style
  5. It offers no valuable insight

My suspicion is that problems 1 and 2 can be corrected. Possibly problem 3 can be partially corrected, though that would seem to require a massive cognitive leap or increase in meaningful memory. In other words, maybe you can teach an AI never to use a cliche, an empty phrase, a redundancy, or a long word when a short one will do. But a Google-sized content index would be expensive.

And in any case, taking away the crap doesn’t give you an original writing style – problem 4. It just makes you write like a less annoying robot. 

And even if it did, so what? As a business owner, you don’t need to worry about “writing well”, let alone in an original style. Sure, that helps. But all you need to stand out from the millions of machine and human content-makers is to provide some valuable insight.

Un-robotically yours,
Rowan
 

PS. I understand that if the singularity happens, the concern expressed in this email may become moot. “What if AI treats us the way we treat less intelligent animals?”