Toning Down the Discourse

Transcript

Joel Goodman: Tired of all the AI slop leaders? This is Higher Ed Hot Takes. And I'm Joel Goodman.

I am delighted to be joined this time by my good friend Shane Baglini.

Shane Baglini: Hey, Joel. Good to see you

Joel Goodman: AI has infiltrated everything that we do.

Shane Baglini: It has.

Joel Goodman: In higher education and we've been back channeling a bit over the weeks as we see things that pop up in our LinkedIns and every other place online and thought it might be good to just have a conversation from the viewpoint of someone that is inside a college, you know, has worked inside institutions for years.

While I've been working outside of institutions and maybe just bring a little bit more grounding to this conversation around how we can actually use large [00:01:00] language models in this technology in ways that are helpful and not harmful. And maybe just call out some of those harmful areas that we think we're seeing across the industry.

'cause there's, there's a lot. Um. This was kind of initially kicked off by a, a couple of different comments we saw online. Um, some very hot takes. We've seen from friends, from colleagues, from vendors that are all the things I've been talking about on this show for the last several, several months. I.

Shane Baglini: My first take is this conversation might come off As you know, we're against ai. Not the case. I, I use it every day. I've used it today. I'll use it tomorrow. I think it's incredibly useful and helpful and a great resource and tool that everybody should, should use and should experiment with and see, see what they figure out.

I, I mean, I'm still finding new ways to use it. I'm still finding new things that it can do and the possibilities are are great. I think the thing that I would take issue with, and I think one of the problems that higher [00:02:00] ed practitioners have is this idea that if you're not advanced in AI or if you're not using AI assistance or don't have an AI chat bot, or if you haven't optimized your entire website, you have for AI search, that you're somehow behind the sort of the, the onslaught of ai.

Quote unquote thought leadership is this sort of, um, and I'm, it's not purposely making people feel that way, but it, it. It does, and I see it all the time on LinkedIn. I hear it from colleagues, you know, I feel like I'm behind on ai. I see all these people doing these great things, and the reality of it is higher ed practitioners, including myself, simply don't have the time to do these things.

And so hearing from people that are genuinely thought leaders, both in and outside of higher ed, it sort of is, is wearing on me this, this constant, you need to be using AI in this way. You can do this with AI and, and really there's. Only a few practical things that I see throughout the week that actually [00:03:00] would work for higher ed practitioners.

And so that's kind of where you and I have been chatting about this, that, you know, it's great, but how does higher ed implement it and where do we get the time to do it?

Joel Goodman: I think there's, there's a multi-layered issue, right? I think there's been so much hype, even just outside of our own circles, around the ways to use this technology. A lot of that hype is unfounded. A lot of that hype is really just to sell. Subscriptions. A lot of that hype is, I think, detrimental to the quality levels of the, of the work that we produce and the output that we have.

I think it's detrimental to parts of society, and I've talked about, talked about that on this show. And then there's the follow the leader style that higher Ed seems to have and has always had, where the thought leaders, or as I've been calling them, the AI slop leaders within the industry, it was within our industry, within a lot of industry.

They're following those hype cycles, right? And then part of that is because you've got people that are just kinda like cheerleaders for it, regardless. I don't [00:04:00] know what their motivations are, but they're cheerleaders. Other people are just making money on it. Usually those are the vendors, usually those are the founders of these companies that are putting the hard sell on institutions and convincing them that they have a solution for a problem that does not exist, but telling people they have this problem, right?

And that constant spin and cycle of hype and pressure and filling of any empty space in the, in the discourse and the conversation with this information, I think does put that. Downward pressure on people to think, oh, I have to do this, which is really strange to me for an industry that has traditionally been so risk averse that they would think that jumping into a technology that is as young as it is with as many faults as it has, the, the amount of misinformation that gets generated by large language models, the, you know, the so-called hallucinations, which.

Really hallucinations is just how it's programmed to do it. You [00:05:00] know, the mistakes that are made in that, and then the lack of criticality on the side of a lot of people that are using these tools, they don't, you know, they think, well, yeah, it seems good to me, and especially if you don't. Know how to course correct for something like that, right?

Like how do you fix that? Caused this extremely traditional industry that we've worked in for years to all of a sudden abandon all of the criticality and thoughtfulness and careful and just start throwing money at it and, and I don't know if that's due to all of the pressure surrounding it. Maybe everyone's out of ideas.

Maybe no one really feels like they know how to do their job anyway. And so like that's, that's what the change is. But those are the things that worry me because we're replacing work that we know works. The thoughtfulness and the expertise that we have, that our colleagues have with something that's supposed to automate it, but we're automating the wrong parts, in my [00:06:00] opinion.

We're automating the stuff that like we as humans should still be working on.

Shane Baglini: I think, I think there's a lot that AI can, can do for us. So two things on what you just said, the, the, the things that humans can do. Right. You and I were going back and forth on this, the CEO of perplexity talking about their browser taking over recruiter jobs. And when I read the headline, I mean, I read the article that the headline to me was just, I'm not a recruiter, obviously. It, it's just insulting to the profession of a recruiter. You know, my current position at William and Mary, I worked with a recruiter. She was outstanding. She was the best recruiter I've ever worked with. She had insights into the organization that were incredibly helpful, including, you know, things like personality fit, institutional fit, and to say that your browser can replace what this recruiter did for me in that process is insane.

There's a lot of uses for ai. That's not one of them like that is

Joel Goodman: Yeah. Well, and it's a huge [00:07:00] overstatement too, right? It's, it's an overstatement to think that a flawed software program is better than flawed humans. At least humans can course correct. The flawed software is programmed to do that sort of thing. And so like I, this has been bugging me like it's because it shows up.

This is a, I think this is a great example of it. Like we are so content to get 70 or 80% of the way there with software when before that would've been inexcusable. If we as ourselves, were only doing 70 or 80% to progress our institutions to progress, our marketing, to progress, whatever. We'd be fired. Like that's, you know, it's fine 'cause computers are doing it.

Um, but no one backfills the other 20% in most cases. I mean, sometimes people do, but like in most cases like that, other 20 percent's not getting backfilled. And when you're trying to automate things like a recruiter's job that requires soft skills, right? Like it requires those interpersonal skills that you really can't, [00:08:00] you can't get out of a predictive text model.

Shane Baglini: No, and it, you know, I think it unearths a bigger problem, um, website as a second impression. And, you know, these AI like chatbots and things like that, that are overlaying your already existing websites and things like that. And, you know, the idea of this recruiter perplexity situation, it's like scanning LinkedIn for qualified candidates is not replacing a recruiter.

It's a tool for a recruiter. The same way. AI is a tool for our websites, for our marketing, for our messaging. The, the bigger issue that's getting covered up and we're trying to bandaid with AI is that most of our websites stink and the information is not good. And so asking AI to correct that or using AI to mask that is exacerbating your issue. If you're reliant on AI search and then students get to your website with maybe higher intent, who knows? And your website stinks.

Joel Goodman: Yeah.

Shane Baglini: It doesn't matter if you showed up in the search results. It it, it [00:09:00] doesn't matter because if the student can't find what they want when they get there, they're going to say, okay, great.

That came up in the AI results. Let me move on to the next institution.

Joel Goodman: Right. If your content's bad, if your website's not organized properly, that 80% that the a that the AI tool gets you to helping them decreases, uh, you know, it's a, it's a garbage and garbage out sort of deal. To your point around man AI site searching, I've been having this conversation with a lot of people that are producing these products right now.

I'm like, AI site search. Or conversational site searches. A lot of people are calling it on websites, I think again, gets you a certain way there. But remember when all those snippets started showing up in Google searches originally, and we were all like, this sucks. This is awful. This is a terrible experience.

But then we put 'em on our websites like we're, you know, and to your point, it's like we're serving up the information, we're summarizing something on the website, which one means like the content on our website must not be [00:10:00] very important. And so like. Maybe question how much content's on your website if a summary is enough to give people what they need.

But two, those summaries don't lead to actions. They answer a question. The person leaves your website, they may never come back, you know, and so you're leaving a lot to chance. It's like kind of the exact opposite. Conversion rate optimization practices. If you're going to summarize something or provide an answer, you need to provide an action as a next step for someone to do.

And these tools don't do that yet. It's like a huge oversight by the producers of these, of these site search tools. And I keep telling them like. People don't come to your university website or your college website to just be informed. And if they are doing that, like you're doing your job wrong because you need to turn them from people that are learning about your institution to people that are interacting and building affinity with your institution.

And that means getting them to take actions and connect with you, but they don't, you know, that gets glossed over. And unless you're a UX designer, unless you're someone that is deep in the weeds of how this stuff works. In a [00:11:00] web interface, you're probably not thinking about it when you get that sales pitch.

I, I think it, it just highlights one of the, one of the missing pieces to this, put AI in everything at all costs, mentality that we see just cropping up in culture.

Shane Baglini: And I think another point on the. Google search AI stuff is that, go look at the sources of most of the search results coming up, and what you're gonna find is Reddit and Quora and other public sites that you can't control the information on. So if your website is not in order, if the experience that students are having when they get to your website is a bad one, it only compounds the problem because there's gonna be information that this AI is pulling from that you can't control.

What you can control is the user experience. Once somebody gets to your website, how you're leading them to some sort of conversion action. The messaging of your brand, the, the usefulness of the materials that you're providing people, those are things that [00:12:00] AI at the moment can't do. It's on you to sort of build the foundation and then see how AI can enhance it rather than understanding that we've got a problem.

Maybe our website is old, maybe there's too, you know, it's become a repository type of situation. Let's have AI fix it by having us show up and search results is not solving your core problem.

Joel Goodman: Yeah, I, I think if anything, this sort of underscores the systemic problems we've had in web marketing, digital marketing across the, the sector forever, ever. And if you know those summaries are drawing from sources that are not your website or are not people that. Have had good experiences with you or have not had a positive interaction or you know, a reason to talk like good about your institution.

That's not an AI problem. You know, an AI's not gonna fix it. Like that's a systemic problem within how you're operating as a brand, how you're operating your student [00:13:00] experience, and touch points that people have as they come into your institution. I fear. That despite that being the truth, the answer still becomes like, yes, but AI will fix.

All of those things for you. And then the question is, how does it fix it lie for you? Like, is it, is it going to make things up because that is what it does and that that becomes, you know, a bigger issue and, and it becomes a bigger issue That is in line with how the slop economy kind of works in general.

You know, when you're seeing images and videos and you can't tell if they're generated by a model or if they're real. Slop. If you're reading content that doesn't quite make sense or you can't tell if you know if it's supposed to be that generic or not. Like that's slop. If everything that we're doing is just kind of regurgitating and cycling through the same models over and over and over again, that's just building up on lop and it presents a real opportunity for those that in this industry want to take it.

And that's to [00:14:00] be human again. That's to. Have a point of view that is not the median of all of the content that has been used to train these models. Set yourself apart. You don't have to play into it. Going back to your earlier point, if you feel behind, you don't have to play into this like. Think critically when you can about how to use these tools.

Where I use these tools is for analysis. I think it's good for data analysis. It's good for finding patterns in what you're doing. You know, if you wanted to do some information architecture work on your website, I. Go and do a screaming Frog poll of your entire thing. Take that CSV and give it to a model and have it analyze it and give you some ideas about what it sees.

There are things that can give you superpowers as someone that is an expert if you use and leverage these tools in the right way, because they are just machine learning and that technology has been around for a long time.

Shane Baglini: Yeah, agreed. That's a great point. Using the [00:15:00] tools as ways to expand your skillset, I think is, is one of the messages that I'd like to get across to people. I'll never, ever claim to be a mathematician or remotely good with numbers, and so for me to have a tool to be able to do financial analysis and do cost per acquisition modeling and things of that nature, that's invaluable to me.

That's where AI is like the best thing ever because it's letting me do things that I can't do. Based on just who I am and how I'm wired. The other thing too is that I've found that a pretty large portion of my AI use is using it to solve the things that take mental energy, but aren't a priority for me.

So, you know, it could, it could very well be like drafting, I dunno. Thank you. Page language. After your RFI form, could I do it in, in a half an hour? I could. Could I do it in 30 seconds and get 80% of the way there? I, I could, and that's saving me mental energy to [00:16:00] focus on other stuff, leading people, teamwork dynamics, thi things again, that AI can't replace that I need the, the brain bandwidth to do.

So I use it every day on things like that, on things that expand my skillset, but I, I can't use it to mask the fact that. I even have to check the numbers and I'm not good at math, so like you can't use it as a, a, a, a, be all, end all. You can't use it as a bandaid for things that you can't do or that you're, you know, your brand can't do, your messaging can't do there.

Those are foundational things that AI can, that enhance once you start to work on, on building that foundation.

Joel Goodman: And I think it's key to understand what this technology is best at. Kind of like put it in tears, right? It's like financial analysis. You know what it's really good at? You could have it just write a script and have it execute the script and then you know the script is gonna run the exact same time every time.

It's going to give you that data consistently. Whenever you need that to happen, you plug in your numbers and let [00:17:00] it do it. Versus like assuming that when you go and ask it to, to conduct an analysis, it's going to do that. Like have it write a script, it'll do it, and then. You can execute it with, with the lm you can execute it on your own if you learn how to get in the command line or whatever.

But like, there are ways to make sure that it's doing things consistently and it's doing things that you spend less server cycles, you spend less energy, um, you know, just from a, from a sheer processing standpoint. If it's writing a script once, rather than trying to write a script internally, maybe slightly different every single time and then run the thing, you know, and it's the same thing with, uh, it's kind of like with prompt engineering for, for different, uh, generative outputs.

You know, if you want it to be consistent, you write a specific prompt and you make sure that prompt gives it some guidelines to do that, and it will generate that text consistently each time finding the ways to. Put those guardrails in place for things based on what large language models are actually good at [00:18:00] versus what you're told they're good at, or what the marketing has sold us.

Because generative text, I think, is one of the most boring things that LLMs do. Like, I mean, it's cool, it's very cool, but it also is so prone to error where writing, writing code. In small subsets like full apps, that's a different, then we can talk about that some other time. But like small scripts, automating small things, like making sure that certain routines are in place.

Those are the things that it's very, very good at doing those, doing that analysis, writing a script to do that analysis, you know, that sort of thing. So using it to do the things. It's really good. Um, and also kind of like cementing it so that it's not changing that in the future as models change, but it's a skill A lot of people don't have.

What, because maybe you're not technical. Maybe you know, you, you haven't looked into it enough like, but there are things that, that large language models are a lot better at doing than other things. And they're typically not the ones that we've [00:19:00] been sold.

Shane Baglini: I think for higher ed and probably more industries, we have so many things going on at one time and so many initiatives that we're trying to push forward. I think a lot of times innovating with AI gets left behind because of deadlines, because of KPIs, because of things that we're responsible for. And so I would just, again, drive home the message that if you're seeing a lot about AI and you think, I'm not, I'm not doing any of this.

I'm behind, I, I, I gotta catch up. You're right where you should be because. You are focusing on other things. You're focusing on things related to the mission of your institution. You're focused on helping students. You're focused on engaging with donors. Whatever area of higher ed you're working in that you think I need to catch up on AI to do my job better, it's probably not the case because the things that you're doing, building relationships, displaying emotional intelligence, you know, leadership qualities, [00:20:00] team building.

All of those things that you focus on can't be replaced by ai, and therefore you can't be replaced by AI if you don't, if you don't think you're up to speed. So I would just enc, I just encourage everybody to chill out, relax, do it at your own pace. If your organization or your institution can allow you to block two hours on a Friday afternoon to experiment with ai, if you have it in your schedule, that's great.

But a lot of the pain points I hear from colleagues around the industry is that I don't have time to do this stuff. We are burning the candle at both ends. We're short staffed, short on resources. I just don't have time for this. And so that's where I feel like this kind of pressure, this pressure complex is coming from that, that I, you know.

Constant ai, onslaught of thought leadership and hot takes and all this, this stuff about ai, it's, it, it, it, it can make you feel that way and I'm just would encourage people to not feel that way 'cause [00:21:00] I'm in the same boat as you. Everybody's in the same boat. And I think it, the reality of what higher ed marketers and higher ed in general do and go through on a daily basis is being sort of pushed aside for AI implementation.

Joel Goodman: If you need help figuring out how to use these technologies, there are plenty of people that would love to help you do it. Get in touch with me, subscribe to the newsletter, and I'll be happy to connect you with folks or help you myself if that's something I can do.