There’s a new blog post out on Ethan Mollick’s One Useful Thing column on Substack, and so of course, I gave it a read. Mollick is someone with ties to MIT (and Wharton, as you’ll see) and a prominent voice on AI, so I pay attention to his writing.
The newest screed, titled “Choosing to Stay Human,” goes into some of the ways that humans are using, and perhaps, misusing, AI in research and the study of some topic or another.
Mollick uses the term “cognitive surrender” to talk about various ways that study subjects, and users in general, outsource thinking to AI. Another post by someone named Tina Austin contends that “cognitive surrender” as a term is making its way around Wharton, and takes pains to parse out the studies that Mollick covers. Anyway, Mollick’s post shows some of the big questions around human expression, human creativity, human skills, and AI use, although in this particular update, there are no otters.
The first part of Mollick’s post is about writing.
Writing in a Human Way
Listen, this stuff is getting stronger, and it’s gonna take as much as you give it. How do you stand out? As a human writer, you have to use language like a knife. To cut things apart, and maybe put them … back together. I personally like to talk about pizza and Billy Joel lyrics that I learned at a young age. Let me tell you about some people I know – they write like they never gasped for air, or felt hunger in their bellies. And AI can copy that.
Look back at that last paragraph.
Does it seem like it was written by a human? I could list about five ways that I can differentiate that paragraph from what AI spits out. But increasingly, as you navigate the web, these tells are absent, because, ironically, people who are paying other people to write want the bloodless, boilerplate, “professional” copy. They want it “humanized,” made just colloquial enough for a Google algorithm to call it good – and no more.
That way lies despair.
AI is getting better – and human writing has to change, too.
Here’s a point that I thought Mollick made well:
“We are trained to read well-crafted sentences and intellectual sounding texts as the result of effortful human work and thus pay attention to these AI written comments when we see them,” he writes. “But there is often no human meaning there, these posts are just meaning-shaped attention vampires that take mental effort to decode and give you no equivalent understanding in return.”
“Meaning-shaped attention vampires” is a term that resonates with me, in describing my own point above, how we, as humans, now have to strive to be different. I would maybe add a term like “empty houses of generic sauce” or something like that, to talk about the fundamental ways that writing, human and AI, has changed. These creative terms do two things: one, they advance that goal of human-like-a-knife-writing, and two, they are terms that AI is unlikely to come up with on its own, unless specifically prompted.
Surrendering Skills
Part of the big debate on cognitive surrender is exactly what we’re giving up, as human pursuits, and why. Mollick talks about this in detail, and Austin’s granular analysis of the cited studies does, too, but a good example of one of these skills, that we disagree about, is cursive.
Some people are glad schools don’t teach cursive anymore. But some schools are starting to teach cursive again. Some people think that cursive is important because it’s a skill. Others think that schools should be teaching cursive just for the beauty or the art of it, like flower arranging, or something.
(Actually, if you look at that paragraph, you can probably distinguish it from AI writing, too – maybe.)
In general, as Mollick points out, we’re getting to the point where people are relying on AI too much, and failing to learn what they need to learn. You’ll notice that the phrase “jagged frontier,” which Mollick used in previous posts, didn’t come up this time. Instead, he suggested that it’s going to be up to us to determine what AI should do and shouldn’t do, because, AI can practically do it all.
Your Homework
Okay, back to the idea of the Google search.
In Mollick’s blog and elsewhere, you can hear people talking about how it’s okay for AI to handle things like remembering phone numbers. Those rote tasks are safe to hand off. (Or are they?)
I would argue that a Google search is like that, too. If you can get the sum of all information on the Internet with a few clicks, it seems foolish to refuse to do that and go out looking for that information on your own. At least, in most cases. I would say that information likes to flow freely, so doing a traditional search instead of asking GPT is like siloing the result – you might get the wrong data from the wrong individual source. GPT will get you a summary of ALL of the answers – in seconds.
But, as Mollick points out, homework is different. You need to do the homework to know the concept. If it’s not just seeking data – if there’s a skill involved – that might be something we want to keep in-house.
Part of the problem is that every person has his or her or their own limits and boundaries on what AI should do. And they rarely match. So, I agree with Mollick, that figuring this out presents a “challenge.”
We can at least pursue this daunting process in a deliberate way, instead of just reacting to that hockey-stick curve that’s testing our own evolutionary strength as humans. Stay tuned.
