Craft amongst the slop

Transcript

As a web designer and marketer, I've seen it all and unfortunately, it all looks the same. You've heard this though, right? I've been in the higher ed web marketing game since 2009 and then, as now, we were talking about how everyone's website looks the same. Or how every institution has some version of the same tagline.

One of my favorite videos is of my good friend Jeff Stevens doing a dramatic reading of dozens of university taglines to underscore how similar they all are.

So what happens now that we have generative AI and LLMs? A lot of us have been integrating ChatGPT into our work, having it draft emails or website content or news stories. And that's not necessarily bad if you have talented writers and editors who go through and revise those drafts. But I suspect not many of us do... or not many of us want to wait around for an edit.

And when budget pressures come around, as they always do in higher ed, we turn to cutting staff and underfunding marketing. Especially now with the promise of automation through AI, why can't we just replace that writer with ChatGPT? It seems good enough.

Or how about decreasing your frontline admissions staff because you can replace them with a chatbot? Even creepier, replace them with multimodal AI that synthesizes voices and can talk on the phone. That stuff freaks me out.

The hitch is that the stuff that we have right now is not actually AI; it's machine learning. We've had machine learning for 20-30 years available to us in Higher Ed and plenty of more accessible ML tools have been in our toolbox for the past 10 years or so.

It's a fact that generative AI, all of which are trained on the same core data, can't actually understand anything — they don't think. LLMs are predictive algorithms. They guess what word goes next to other words. I haven't seen any proof that it knows how to differentiate, say, Wheaton College in Illinois from Wheaton College in Massachusetts. At least not to a degree that ensures you are not giving prospective students incorrect information or information that contradicts your values and brand.

LLMs are, at minimum, a decade away from anything like real thought and interaction. Right now, their responses are purely predictive, and while they might trick you into believing they know things at times, we've all experienced the hallucinations. It's giving too much credit to say they make mistakes because these algorithms have no actual knowledge or thought. They don't make mistakes, they just select inaccurate information to serve up.

Like, I just think there are way too many questions around the ethics of that for us to blindly be pouring cash into new tech.

But this comes down to an extremely nihilistic outlook on society. A lot of the uses for AI being sold to us, across all sectors, are just cynicism wrapped up in a sales pitch.

When you outsource your hiring to an LLM, what sort of respect are you showing your applicants? How would you feel if you found out your new boss didn't value your expertise enough to write an acceptance letter?

When putting an AI-powered voice bot between your admissions staff and prospective students, do you feel you're showing real care? These are the things that keep me up at night. We're losing authenticity, we're losing connection, we're taking humanity out of inherently human experiences.

And when you take the humanity out of the connection with your prospective students, you lose the edge you have. There aren't that many differentiators in the marketing that HigherEd does, that when you lose that core one—your people and your culture and the ethos behind what you do—and give it to an algorithm trained on generic slop, you lose — and so do your prospects.

Your people are the one unique selling proposition that you have over another institution. If you're shoveling your prospective students off to talk to a robot, essentially, there’s a question about how much you care about them and the process. How much do you really trust a large language model or a generative model to know your brand and to accurately express your values?

I'm not saying this level of disrespect is intentional. Most of us in higher ed are good people. But the level of disrespect that is borne out of a lack of serious consideration is noticeable. If the quality of your content declines because you didn't care enough to put an actual talented person on the job, you're disrespecting your audiences. When your GenAI images start looking like everyone else's, you lose market share and brand standing while infringing on the work of real artists.

Slop begets slop. And a resurgence of craft is the only remedy. The web is now filled with AI-generated slop, and the race to hegemonic mediocrity is full sprint. Higher ed already has a serious trust issue, and that's only going to get worse if we don't remember that humanity is at the heart of the work that we do.

Look, LLMs have real uses, and they are largely the same as what we've used Machine Learning for for years. So think a bit broader. Generative AI — the stuff ChatGPT, Claude, and Midjourney do — is the lowest common denominator and the easiest thing to sell. It's surface-level stuff and honestly, the stuff a real person should be doing in production.

You should be using LLMs to support your craft, not replace it. That's the lesson here. Use them for research or data analysis. Use them for finding patterns or connecting your different databases together. Use them to manage data integrity or keep your notes organized.

But don't outsource your human touch. The world doesn't need more slop. You're better than that.

Thanks for listening. If you want to connect, find me, Joel Goodman, on Bluesky and LinkedIn.