How You Can Tell This Was Not Written With AI

It was written with my beta version of HEW (Hard Earned Wisdom)

Image by Sasha Moroz from Pixabay

Here’s what I did not do as I was writing this piece

I did not fire up ChatGPT, Copilot, Gemini, Claude or any of the LLMs lurking on the Interwebs. I did not spend time composing clever prompts to elicit an essay I could pass off as my own.

Grammarly did chime in while I was writing with its annoying, color-coded underlining because I don’t know how to make it shut up (and that sentence alone got me two underlines). I keep trying to turn it off, but it keeps coming back. It’s too smart for me.

This is because LLMs (Large Language Models, in case you’ve been off the grid and camping in your backyard for the past five years), even modest ones like Grammarly, are very, very good at what they do.

The largest LLMs have read and been trained on more text than you or I could read in 20,000 years. And the LLMs, for lack of a less anthropomorphic term, remember all of it, in that they can access it in nanoseconds.

I can’t remember half of what I read two months ago, let alone 20,000 years ago — oh, wait.

LLMs don’t just parrot back rote text when prompted. They’re able to synthesize, choose, and modulate responses. Not just in written text or digital images, either. They can generate creepily authentic-sounding human voices.

Recently, my husband took advantage of the free beta-testing of Notebook LM. He entered the entire text of a memoir he’d written and asked Notebook to critique it using certain criteria.

The results were astonishing. With few exceptions, Notebook was able to summarize his memoir with stunning accuracy, as well as identify its themes better than most book reviewers on Goodreads or Amazon.

As if that wasn’t enough, my husband then asked it to produce a ten-minute “podcast” based on his story.

Within minutes, we were listening to what I could have sworn were two actual human beings, a man and a woman, talking about the memoir as though it was part of a series of book reviews. They laughed, chatted, shared impressions of the book, and sometimes questioned each other’s interpretations, with jaw-dropping verisimilitude.

These disembodied, digital podcasters were so convincing that in the very few instances when they conflated elements of the story incorrectly or were off base in what they said about one of its characters, we were offended.

“How could they think that?” we protested. “Were they reading the same book?”

Until we remembered this wasn't a real podcast and those weren’t real readers, or people, or anything except products of a highly sophisticated predictive language model.

Which is very, very smart.

Judging by how jaw-droppingly far AI has come in the past two years since the first iteration of ChatGPT was launched, its abilities are likely to increase exponentially.

AI already displays more “I” than most humans do. Who knows where it will go from here? By this time next year, even the dimmest human bean with a working smartphone may have access to all of the information humanity has ever recorded.

That’ll make cheating on term papers even easier. But it will not save us

I can remember those innocent days when we were sure that humanity was boldly going, with Star-Trek-style elan, towards ever-expanding freedom, prosperity, and understanding.

Certainly, the explosion in access to information that took off when the Internet became publicly available would accelerate this process, right?

Remember how the Information Superhighway was going to make us all better informed? Better able to make decisions for the greater good?

Yeah, good times.

But not so much.

Either the world is in a bigger mess than it’s ever been before, or it looks that way because we can all watch it going haywire in real-time, all the time.

It’s not because we lack intelligence. We lack wisdom.

There is no question that our intelligence, our cleverness, our ability to manipulate objects and forces, has given us a great deal. Even someone living in the most modest home in a developed country has access to comforts that medieval royalty would envy: light at the touch of a button, central heating, manure-free transportation, and the luxury of never worrying about smallpox.

But all of our smarts have not kept us out of wars, or from exploiting our fellow humans, or from bringing ourselves closer to the brink of environmental collapse.

We may live longer and have better teeth than we did in, say, the 13th Century. But I don’t know that we’re any happier.

Our discernment hasn’t begun to keep pace with our smarts. When we can do a thing — like split the atom, clone mammals, or engineer artificial intelligence that develops so quickly even its designers aren’t sure what it’s up to — we jump right in and do it.

As a species, we can’t seem to resist the next shiny thing we’ve figured out how to make. Even if what we're making could destroy us.

My point is this: our intelligence is not coming to save us. Not until we develop the wisdom to know how to use it.

And this, I believe, is where we wise old owls come in. There is no substitute for life experience, for surviving disillusionment and emerging with our faith in life intact.

If we’re wise old owls, we know preaching won’t help. Comparing the Now to some distorted version of the past, some illusory Good Old Days, won’t help either.

Maybe the best we can do, as the mad circus spins faster and faster, is watch it unfold — with compassion, without condemnation, but with discernment. Instead of being pulled into the center ring, if we respond rather than react — perhaps with a gentle question: not can we, but should we? — that will be some help.

It can’t hurt to try.