[00:00:04.410] – Kriste Goad

Hey, welcome to How It’s Done, a podcast for curious marketers. I’m Kriste Goad. I’ll be your host, and I’m really glad you’re here. Hello, curious marketers. I am super excited about today’s topic, which is the ethical use of AI technology in marketing and public relations. This is a topic on everyone’s mind. Not a day goes by that I don’t see at least four or five headlines on this subject. And now, as is the case with such things that blast onto the scene and completely disrupt pretty much everything, which is what’s happened. That happened in November 2022, when the introduction of ChatGPT by OpenAI pretty much made generative AI tools widely available for the very first time, which is pretty crazy to think about. That was not long ago. And so now this is a subject that’s being increasingly studied and as we’re starting to see, increasingly regulated, which is why I’m so excited about today’s guest, Todd O’Neill. Todd is an associate professor of Interactive Media as well as coordinator of the Interactive Media degree program for the College of Media and Entertainment at Middle Tennessee State University. Welcome to the show, Todd.

[00:01:32.390] – Todd O’Neill

Hi. Thank you. Thank you for asking me to come on. Appreciate it.

[00:01:35.040] – Kriste Goad

I am so glad to have you here. You’re a wealth of information. I can’t wait to learn more about this topic and how you’re approaching this. Will you tell us a little bit about yourself, a little bit about the program at MTSU and really a little bit about how this topic of AI technology and its ethical use and how’s that playing out on the college campuses right now, in your program, in your thought?

[00:02:01.900] – Todd O’Neill

Yeah, sure. Well, I started my career quite a long time ago in the heyday of corporate video. I was a video producer, but my career has constantly evolved. I was doing multimedia CD-ROM. I did stuff for Prince and Coldwell Banker. I mean, that’s the spectrum of clients I had.

[00:02:25.860] – Kriste Goad

That is quite the spectrum, yeah.

[00:02:28.110] – Todd O’Neill

And then got into the web, which blew my mind about what you could do with the World Wide web. 

[00:02:37.590] – Kriste Goad

Are you an early Adopter?

[00:02:38.950] – Todd O’Neill

Oh, yeah.

[00:02:40.130] – Kriste Goad

Me too. I was on those community bulletin boards back in the day.

[00:02:43.140] – Todd O’Neill

Oh, no. Yeah, no, I go for. I got my first commodore computer in 1985.

[00:02:49.560] – Kriste Goad

Wow.

[00:02:50.630] – Todd O’Neill

Had no disk drive. Like, I would type in these programs out of magazines and then have to unplug my computer from the tv and lose everything.

[00:03:01.610] – Kriste Goad

Wow.

[00:03:02.330] – Todd O’Neill

Okay. Yeah.

[00:03:03.790] – Kriste Goad

You have great perspective on today’s topic.

[00:03:07.370] – Todd O’Neill

Know, it’s, it’s, again, it’s more that evolution of technology and things that have been going on. The program I teach at MTSU is called Interactive Media, and basically what that means is we study and teach online communications of every form, whether it’s web or social or video. And now augmented reality and things like the metaverse and virtual worlds, but also emerging technologies. And that’s how I got connected or interested in AI. It came on the scene. Right. I was doing things with it before ChatGPT came along some of the early, actually another tool from OpenAI called DALL-E, generating images and that sort of thing. And it was relatively. Well, it was pretty primitive. I mean, it wasn’t all that great.

[00:04:07.070] – Kriste Goad

Would you compare it to some of those early days of the Internet?

[00:04:09.970] – Todd O’Neill

Oh, yeah. It’s like Photoshop 1.0 and whatever it is now, Photoshop 27 or something, it’s some sort of crazy number. Yeah, it’s just continually evolved. And that’s what we try to get across to our students, that they need be prepared to pivot when something new comes along and pivot earlier than when it just hits. Right. So, like with AI, our students were working and needed to be working with the early image generation, stuff like DALL-E. Before ChatGPT came because their employers like you. If you have new graduates or interns in your organization, you kind of expect them to know.

[00:05:00.770] – Kriste Goad

Well, I do expect them to know, but then I find that they don’t often know.

[00:05:05.240] – Todd O’Neill

No, yeah, exactly. Because in many cases, academic programs focus on this is not quite the right word, but, like, the basics or the pure foundational stuff.

[00:05:20.870] – Kriste Goad

Right.

[00:05:21.450] – Todd O’Neill

And they don’t really study stuff that might be in the crystal ball, stuff that’s about to happen. Right. Because it hasn’t happened yet. And there’s no data on that, and there’s no research. No one’s written an article about that because it hasn’t happened yet. Well, yeah, but it’s going to happen. And even as academics, we need to be ready for that. We can’t react to it. We have to be the leaders, the thought leaders, to take it forward, and then we’ll go back and study what its impact was or that sort of thing.

[00:05:59.620] – Kriste Goad

So in your course at MTSU, do you have people from across different disciplines that are in your program, or is it pretty much geared toward marketing, communications type professionals?

[00:06:12.830] – Todd O’Neill

Yeah. So that’s maybe the interesting thing. So we’re in the college of media entertainment, and about half of our college is recording industry. It’s about a thousand students who study music business and audio production. We have a whole department that does just video and film production and we have a whole journalism department. We’re the only program that talks about online communication and emerging media. And because all communication is online, all communication is on the Internet. We actually touch on lots of different things. So, yeah, we teach about marketing, we teach about design, we teach about public relations. If you’re familiar with the PESSO model, right.

[00:07:03.240] – Kriste Goad

Yes.

[00:07:03.830] – Todd O’Neill

Paid, earned, syndicator, subscribed and owned. So we teach that model. But we’re not the marketing department. There’s a whole college of business that has actually a program called digital marketing.

[00:07:19.110] – Kriste Goad

Right.

[00:07:19.920] – Todd O’Neill

We’ve been around since 2000. That program has been around for three years.

[00:07:25.150] – Kriste Goad

Oh, okay.

[00:07:26.400] – Todd O’Neill

Yeah, we are at the tip of the spear. We are at the bleeding edge as much as we possibly can.

[00:07:37.800] – Kriste Goad

I love that

[00:07:38.850] – Todd O’Neill

Because at some point, well, like I said, I don’t believe communicators can follow. I think they need to anticipate on behalf of their clients, like, what’s coming next. Because competitively, if you wait and react to something new that’s coming down the line, well, that means your client is at a disadvantage because it takes you time to get up to speed and then to get them up to speed. And if their competitor is working with somebody else who’s already up to speed, well, you’re behind.

[00:08:19.040] – Kriste Goad

Yeah. That makes me think about several years ago, like sort of early 2000s or 2010s. It was like social media was the thing. And it’s like, no, hey, clients, you need to start really looking at this. This is where it’s headed. This is where people are. And they’re like, we’re a business. Why do we need social media?

[00:08:49.840] – Todd O’Neill

Yeah.

[00:08:50.750] – Kriste Goad

And it was a hard sell, and now you got to have it. And everybody understands that.

[00:08:55.660] – Todd O’Neill

Yeah, no, I remember I lived in San Antonio for quite some time, and I was on the board for the advertising Federation of San Antonio.

[00:09:05.190] – Kriste Goad

Okay.

[00:09:05.680] – Todd O’Neill

And this is about 2009, 2010, that era. And I said, yeah, we need to get a Twitter account. And they look at me like a what account? Facebook. Facebook, that’s for college, you know. No, hold on. We actually need to do this because there’s a lot of people there and you need to figure out how to do it.

[00:09:33.670] – Kriste Goad

And that’s back before it was a paid platform.

[00:09:38.640] – Todd O’Neill

That was well before Facebook was like, oh, yeah. They were still all about, come share your stuff with your friends and your family, and then know, flipped that a few years later, and now they’re purely an advertising play.

[00:09:56.060] – Kriste Goad

Absolutely.

[00:10:00.290] – Todd O’Neill

As communicators, we have to face the fact that, yeah, the Internet is the communication channel. Television has moved to the Internet. Audio, music has moved to the Internet. It just happened. It has happened. It’s done. The other things are going to have to ride on the Internet, and we’re going to have to know how to use it and utilize it to our benefit, to promote our businesses and to our clients and customers and employers benefits.

[00:10:32.570] – Kriste Goad

Absolutely.

[00:10:33.480] – Todd O’Neill

I mean, that’s our job, right? We’re professional communicators.

[00:10:37.310] – Kriste Goad

Great point. I need to touch base with you more often.

[00:10:42.350] – Kriste Goad

Well, so in what ways do you believe AI technologies are really reshaping the landscape of PR specifically? And I’m asking about PR specifically because I wanted to hit on a report that PRSA, Public Relations Society of America issued around AI. And they’re one of the few, you know sort of professional organizations in PR, marketing space that has done so so far. And that’s a part of what we do here at fuoco is PR. We’re an integrated marketing firm, but PR is a big part of that. So I’m just curious how you think AI is really reshaping the PR. We can talk specifically about marketing, too. And then what are potential benefits do you think they offer for us as practitioners and our clients? To your point.

[00:11:37.470] – Todd O’Neill

It’s important, I think, to separate the technology from the products. Right. So the concept of large language models, which is what open AI bases ChatGPT on, right?

[00:11:52.280] – Kriste Goad

Yes.

[00:11:52.850] – Todd O’Neill

ChatGPT is a large language model. That’s a product. Right. But let’s talk about the technology of things like large language models. So the technology, super exciting, right. The things you could do with your, just internally go and analyze massive amounts of data. Imagine being able to say, give me the, this is something you spend a lot of time to research. Show me the trends in a chart that is purple of the number of impressions over the last ten years on a particular demographic. Just type that in and a tool will generate that for you if you have all the data about that. Right, and that can be your data, and that’s really important to remember. It’s your data, your information, that’s incredibly powerful. And we’ve never been able to do that. And if someone could argue and say, well, yeah, we created that chart three years ago, how long did it take you to compile the data, like actually find it? Well, it took us like two weeks to go. No, I’m talking about something that can happen, like in a minute.

[00:13:27.970] – Kriste Goad

Yeah, that’s crazy.

[00:13:30.290] – Todd O’Neill

That’s insane. And so productivity opportunities crazy. I think it’s a fear. But there’s this whole issue of, oh, I’m going to lose my job because AI is just going to take it. It’s like, no, you didn’t lose your job when spell checkers were introduced. Didn’t happen. Didn’t happen. Will it change your job? It better.

[00:14:03.230] – Kriste Goad

Right. Or you will lose your job. Right.

[00:14:06.990] – Todd O’Neill

Or you will lose your job. Yeah, it better change your job. And it just means you’re going to be able to do your job better and do it faster and do it more interestingly and come up with different and more innovative solutions than you have been able to before because a lot of the grungy kind of work has kind of been pushed aside a little bit. You’re letting a machine do it for you. So, yeah, the opportunities, amazing for professionals and just the technology itself just opens things up. That’s just the technology. And we can talk about the products later if you want, but, yeah, well.

[00:14:49.770] – Kriste Goad

So what do you recommend people do? Just get in there and just start using it. What do you recommend to your students? How do you teach? How do we, any of us? But particularly if you want to do it as part of, as a practitioner, what do I need to do?

[00:15:08.160] – Todd O’Neill

So I would say first, and let’s now talk about the AI products that are out there. Right?

[00:15:15.900] – Kriste Goad

Awesome.

[00:15:16.470] – Todd O’Neill

If you’re going to use those products, understand they are products. They are created by companies whose bottom line is their bottom line to return value to their shareholders. So they’re not giving you stuff for free. You are paying in some way. And I’ll talk about that in a second.

[00:15:42.700] – Kriste Goad

Yes, please.

[00:15:45.790] – Todd O’Neill

The thing to remember with these products is, yes, they’re pretty powerful. But the other part, and if you look in their terms, conditions or in some cases right on their front page, they will tell you they are an experiment.

[00:15:59.970] – Kriste Goad

Right?

[00:16:01.250] – Todd O’Neill

It’s like.

[00:16:03.340] – Kriste Goad

Like you go on OpenAI, it’s like Beta.

[00:16:06.550] – Todd O’Neill

On Anthropic, which does the Claude chat tool. I think it’s right on the front page, down in the bottom left that says this is an experiment. And essentially we may make stuff up and have hallucinations. And you better check it all out before you put it out there. They make no bones about it. So, yeah, you have the product. And the thing to remember is one important anything that you use to tell that tool, let’s just talk about ChatGPT, because it’s words, right? Any words you put in there and any response you get is used to continue to train the model, right? So it’s not like thrown away in the pro version of ChatGPT, I believe you have a way. You basically have a checkbox that says anything I put in, you need to toss, you get rid of it, don’t keep it, don’t use it to train your model. That’s fine. Right.

[00:17:18.510] – Kriste Goad

So you can pay for that privilege.

[00:17:22.030] – Todd O’Neill

You can pay for an experiment to not use your information to continue the experiment. Right.

[00:17:33.010] – Kriste Goad

And basically there’s no way to check that whether or doesn’t happen. Right.

[00:17:38.000] – Todd O’Neill

No way whatsoever.

[00:17:41.250] – Kriste Goad

They just put that in there for the lawyers.

[00:17:44.230] – Todd O’Neill

Yeah, kind of.

[00:17:46.710] – Kriste Goad

Actually, no, it’s a way to get more money.

[00:17:50.790] – Todd O’Neill

Yeah.

[00:17:54.230] – Kriste Goad

They’re going to get mad. Hear me say this, because, you know, they’re listening. They’re going to be listening. Anything?

[00:17:59.770] – Todd O’Neill

Oh, yeah. No, they’re listening to everything. Are you using an AI tool to record this zoom?

[00:18:06.660] – Kriste Goad

I am not.

[00:18:08.350] – Todd O’Neill

I have. They’re a little scary.

[00:18:10.670] – Kriste Goad

They are.

[00:18:12.430] – Todd O’Neill

And you don’t know where that information is going to. In terms of professionals using the products. They just need to know that any information they put in is going to go into the model. That means every other person who uses the tool may get some element of that information that you put in as part of their response. Totally possible. It’s so possible that recent research has shown that someone used OpenAI’s image generator and was able to almost precisely reproduce stills from feature films.

[00:18:59.070] – Kriste Goad

Wow.

[00:19:00.590] – Todd O’Neill

Based on how they worded the prompt, the New York Times, as part of their legal case right now, was able to get the prompt to return word for word, paragraph for paragraph, a story from the New York Times.

[00:19:15.780] – Kriste Goad

Wow.

[00:19:16.640] – Todd O’Neill

So it wasn’t all mixed up with the data. There was a way to extract it. So as a PR professional. Yeah. Never put your client stuff in there. It all has to be very generic. I would never recommend anybody putting in things like product names, specs, prices, anything that could be like, we have personally identifiable information, Pii, anything that could, like.

[00:19:48.170] – Kriste Goad

Anything that you say, hey, this is confidential, don’t share this with anybody.

[00:19:51.390] – Todd O’Neill

Well, yeah. Well, even stuff that you might have out there public, that is like more detail about your product. Well, number one, if it’s been out on the Internet, it’s already in the model.

[00:20:04.100] – Kriste Goad

Right.

[00:20:05.250] – Todd O’Neill

It’s just there, but, yeah. So any kind of proprietary or confidential information goes nowhere near any of these products that are out there. Never put that in there. If you’re just trying to clean up a generic paragraph or even an article. Yeah. Great. Yeah. You can do that and it will do a great job of cleanup. If you need to generate ideas. This is one of the things it’s really good for is doing ideation, where you say, hey, I need some ideas for blog posts for my client. Without naming the client who’s in this industry and markets these kinds of products without naming products. Right. To establish like social proof or brand credibility. I’m creating some articles about the industry. So, yeah, it’ll come up with five ideas in a second. We did something that maybe I shouldn’t say, but I will. This is the worst case, right? We are in a meeting to create a mission statement. So if you’ve ever been there.

[00:21:19.310] – Kriste Goad

I’ve been there. Yeah.

[00:21:21.770] – Todd O’Neill

A room. Twelve people trying to create a mission statement. Insane.

[00:21:25.430] – Kriste Goad

It can take a lot longer than one could ever imagine.

[00:21:28.030] – Todd O’Neill

Yes. So ChatGPT is out there. I play around. So I’m in the meeting with my computer and I go create a mission statement for a department at a school in Tennessee that teaches these kinds of things. And it should be two sentences, separate paragraphs. And it goes, and it was not bad. And someone sitting next to me was writing her own piece. I said, well, send that to me. She did. And I said, and it was kind of long, right? Because mission statements are supposed to be compact. So I just said, put this in here, shorten this, put it into simpler language and make it about, I don’t know, three brief sentences long. I believe we really should use that mission statement because it was excellent.

[00:22:21.370] – Kriste Goad

That’s a great example.

[00:22:25.790] – Todd O’Neill

You can do amazing stuff with this technology, but you got to remember it’s product, it’s just like a Facebook. You give Facebook all your information about who you are and it gives you.

[00:22:41.780] – Kriste Goad

And they then own it.

[00:22:44.470] – Todd O’Neill

They don’t necessarily own it because Facebook, like an image you put into Facebook, they actually, in their terms conditions, they don’t say they own it, but you essentially relinquish all rights to it for them to do whatever they want to with it. Right?

[00:23:02.460] – Kriste Goad

Including, I hear people all the time, someone has hacked their account or whatever, and then they lose access to all their photos for the last 15 years.

[00:23:08.260] – Todd O’Neill

Facebook just gobbled up almost 2 billion images that had public permission on them. So you didn’t lock it down to your friends or family to create their own image generator for Facebook, and no one can say anything about it because you gave them permission to do anything they want. So this is the thing with these tools, is if you actually read the terms conditions and no one does, you give them permission to do anything they want and you get cool benefits. Yes. But the future really is, and I’m surprised there aren’t more companies doing this. The future really is vertically integrated or private label large language models, or ChatGPTs. ChatGPT is the name for OpenAI’s large language model.

[00:24:10.630] – Kriste Goad

Right.

[00:24:11.650] – Todd O’Neill

But private custom large language models that you could put your own information into and query yourself to come up with things. Imagine here, like a car company taking all the renderings of all of their vehicles, throwing it into an image generator, and then saying, we need a new hatchback model geared toward this demographic, with seating for five and big enough to tow a trailer, and it would generate all these images that they could use to go into design.

[00:25:01.150] – Kriste Goad

Surely those big car companies and others are in the process of doing this very thing.

[00:25:08.610] – Todd O’Neill

They are already. That’s the thing with AI. AI has been with us for a very long time, right? If you’ve done an airline reservation, you’ve been using AI, right? Intelligence has been around for a long time.

[00:25:21.690] – Kriste Goad

Yeah. And we’re in the healthcare B2B space, right? And so it’s been out there for the last several years, lots of companies coming, bringing AI to the table, doing the kinds. And it’s been part of this, oh, you’re going to get rid of my job? But really it’s doing the things that people can’t even do manually, like you say, looking longitudinally across all the oncology data for XYZ and seeing patterns, and being able to very, almost instantaneously do the work that it would take humans years if ever to do.

[00:26:01.060] – Todd O’Neill

Oh, yeah, no, it’s amazing. And so I think in the agency space, and there are some companies who are doing these, but I think there’s going to be a push into something like the agency space, where someone will create a private label, large language model that you can deploy to your own computers, your own little servers, or you deploy in the cloud, but it’s your space in the cloud, it’s not available to everybody, and you put all your stuff into it. All the stuff that you want there, and it’s secure from other people getting into it. And now you do your own thing to the point where if you’re an agency and you have whatever, ten clients, you would probably have ten large language models, one for each one, because you don’t want.

[00:27:01.980] – Kriste Goad

So that you’re not mixing and mingling your data.

[00:27:03.830] – Todd O’Neill

Exactly. But some great benefits there in terms of creating new reports. Imagine this got to be the bane of. I don’t know if you do these, do you do annual reports. Do you know people who author annual reports?

[00:27:19.150] – Kriste Goad

I do not do annual reports. Yeah, I have led corpcoms inside organizations where we did annual reports. And boy, those are worse than the mission statement creation.

[00:27:35.150] – Todd O’Neill

So blow your mind, right? Just have a ChatGPT do the annual report.

[00:27:39.680] – Kriste Goad

Love that. Yeah.

[00:27:41.320] – Todd O’Neill

You just put all the information in, you know, the voice and tone of the corporation or the organization. It can make the charts for you. It will create the images of the happy employees diligently working at their desks. You could have it do an entire annual report. That is not far fetched.

[00:28:01.510] – Kriste Goad

Well, that is definitely going to take the jobs of all those. If there’s anybody left out there doing annual reports, which there are.

[00:28:08.480] – Todd O’Neill

Yes, there are.

[00:28:10.670] – Kriste Goad

Great idea. Well, I was just going to say, what are some of those other, as those things continue to evolve? You’ve already raised and brought to the floor in this conversation the things that we shouldn’t be doing. But what other ethical considerations should we be mindful of as we’re leveraging the power of this thing? But how do you think about it?

[00:28:40.710] – Todd O’Neill

So this is actually a real life example. I had a student call me last spring. She was almost in tears. She had interned for a marketing firm and had got hired right out of her internship.

[00:28:56.800] – Todd O’Neill

Golden ticket for a student. I know somebody like that. Interned for some, for an organization and gets hired, right? I called it the golden ticket. It’s like what every student should want.

[00:29:09.710] – Kriste Goad

To do, and frankly, that’s what every company wants. What better thing is there? Like an intern was so good, you want them to stay?

[00:29:17.700] – Todd O’Neill

Yeah. She was there for four or five years, worked her way up through the company, was like director of content or content operations. So she was responsible. She had writers underneath her and editors and other content developers creating content for their clients. She got called into her boss’s office, said that she was being laid off, that they were outsourcing all of the content creation for the agency to another kind of agency in Costa Rica that would do all the web content development that her department used to do as well as, and this is super scary. As well as their SEO work. And they were going to use AI. So now follow the chain. The agency in Costa Rica uses AI to create content. Do they tell the marketing agency in the US that they did that? Then? The marketing agency maybe doesn’t know that they used AI to create that particular content right. It gets published to a client’s website,

[00:30:35.080] – Kriste Goad

Oh boy.

[00:30:36.570] – Todd O’Neill

And now information is wrong in there because the AI hallucinated

[00:30:43.790] – Kriste Goad

And nobody was checking it.

[00:30:46.060] – Todd O’Neill

The client comes back to the agency, goes, why? You publish this on our behalf and it’s wrong.

[00:30:53.820] – Todd O’Neill

Do you know what’s going to happen to us, to our image and the marketing agency? What do they do? They go to the other agency and go, what the. Hey, what are you guys doing down there? How’d you make this content? It was wrong. So whose responsibility is it to say, oh, yeah, raise my hand, we used AI to do that. But where does it happen? Because as a client, I would want full disclosure from my agency or for my collaborator, my partner that, yes, they were using AI. They used it for this purpose, this distinct purpose, for this kind of deliverable.

[00:31:37.990] – Kriste Goad

Right.

[00:31:38.860] – Todd O’Neill

And this is why they used it. And here’s our rationale and here’s what we could have done, but we chose not to because it would cost more money, because it would take more time. But somewhere there needs to be, right? That’s an ethical issue. It has to be full disclosure. I hope they don’t catch us. Or, wow. Yeah, we could make a lot more money on this job if we used AI to create the content because it won’t cost as much. And so our margin is going to be like that. It’s like, no.

[00:32:17.710] – Kriste Goad

So from your perspective, as an agency owner, I should be thinking about, okay, yeah, we’re going to leverage AI and here’s how we’re going to do it. Client. And we’re going to do this because we can save money.

[00:32:34.400] – Todd O’Neill

Right?

[00:32:34.900] – Kriste Goad

Because clients are always pushing us on price and we’re getting squeezed so hard on price.

[00:32:43.000] – Todd O’Neill

Oh, yeah, I bet.

[00:32:44.940] – Kriste Goad

So really that’s kind of the only thing left to do is say, okay, well, how can we create content faster, cheaper? So you’re saying we need to make sure the client is aware of, hey, yeah, we can do it for less and here’s how we’re going to do that.

[00:33:03.340] – Todd O’Neill

Well, in my mind it feels that way. Yeah. And like PRSA created their AI usage guidelines.

[00:33:12.070] – Kriste Goad

Yeah, let’s talk about that a little bit. What do you see in there that we should know?

[00:33:16.370] – Todd O’Neill

No, it’s all good. I mean, they do cover all the bases in terms of, don’t put proprietary information up there. Here’s good ways to use it in terms of productivity or workflow. Do inform your employees about how to use it. And all those guidelines are excellent, but I think they almost need to change that into like a guideline template that they could pass down to member agencies to say, you need to inform your clients, you need to have an AI usage guidelines document that you can refer your client to. That explains, here’s how we use it. Here’s why we use it and just lay it out. Now, eventually, we’ll not have to do that. Right down the road, we won’t have to do that. But if you’re going to use a commercial model like OpenAI, you need to tell them. And the other part is you need to do your own due diligence as an agency owner or a communication professional, is you don’t just let it do its thing and then copy and paste.

[00:34:38.150] – Kriste Goad

Right? I mean, that’s part of this PRSA guide as well. It’s like, there’s a lot of misinformation and you got to fact check.

[00:34:47.020] – Todd O’Neill

Yeah. In the early days of OpenAI of the ChatGPT, I’m not sure if it was the first round of ChatGPT in November 2022 or previous to this, but an academic put in a prompt that asked for ten citations about a particular topic. Okay, create a list of ten APA style citations about this topic. And he did. All ten were wrong. All the authors were made up. None of the titles ever existed. The journals could not be found on Earth.

[00:35:29.620] – Kriste Goad

What?

[00:35:30.190] – Todd O’Neill

Nothing was right.

[00:35:31.720] – Kriste Goad

He didn’t check any of them.

[00:35:33.610] – Todd O’Neill

Pardon me?

[00:35:34.320] – Kriste Goad

He didn’t check any of them?

[00:35:35.920] – Todd O’Neill

No, no, he checked them. He checked them. It was an experiment. Basically. He goes, I want to see what this can do. And like, it did nothing.

[00:35:44.660] – Kriste Goad

Oh, he’s the one that found out they were wrong. Okay.

[00:35:47.300] – Todd O’Neill

Oh, yeah. No, he did it as an experiment.

[00:35:49.960] – Kriste Goad

Okay.

[00:35:50.860] – Todd O’Neill

Now, there’s recently a massive national political figures lawyer had some documents submitted to a court, and it was discovered that the legal cases that were cited in the documents were totally made up, didn’t exist. And it turned out, yeah, they did use ChatGPT to create them. Like, what are you doing?

[00:36:19.210] – Kriste Goad

Right.

[00:36:21.770] – Todd O’Neill

If you’re going to use a commercial model, there’s risks, but you have to mitigate them by. If you’re asking for sources. Well, always ask for sources if you’re going to have it create something. You know, find out information about this kind of product or this kind of industry and provide a list of sources for those. And in fact, like Claude from anthropic will tell you at the end of that prompt, we believe these sources are accurate, but you should verify them.

[00:37:02.550] – Kriste Goad

Okay.

[00:37:03.460] – Todd O’Neill

And yeah, you need to do that with everything. So what it does is it cuts down on your creation time, but your fact checking time, because you haven’t done the fact checking or gathered the facts as you’re creating, you’re going to have to add fact checking time to make sure that it is real, that it’s really accurate. Right. So it’s going to shift some things for a bit with the commercial products, but if we can get private label, large language models deployed into companies and agencies, and even for individuals, well, now all that’s changed, because now it’s our stuff. We know what we put in there. It is accurate because we verified it before it went in. And we can create tools from it. We can do what we need to do. And it’s just like you throw your art department, who uses Photoshop, you send them images that the photographer that you hired took, and create whatever, a brochure from those images or something like that. Right. That’s essentially what I’m saying is it.

[00:38:22.310] – Kriste Goad

Seems like a great project for your college.

[00:38:28.290] – Todd O’Neill

No, seriously, we are looking at creating our own large language models. We’re looking at putting a data center here on campus. That is one of the issues with AI.

[00:38:39.150] – Kriste Goad

Yeah, talk to me about that.

[00:38:41.510] – Todd O’Neill

Environmentally, it’s insane. Yeah. People don’t necessarily know this.

[00:38:50.750] – Kriste Goad

In terms of the footprint and the energy, is that what you mean?

[00:38:53.980] – Todd O’Neill

Energy and water, it takes a lot of energy for cooling. Right? So it uses hundreds of thousands of gallons of water every hour for a data center. And it takes hours, not hours, days and weeks to actually train these models. Like the original data set, that ChatGPT used in November 2022 was from September 2021. It took them that whole time to train and refine that model. That’s computers churning the entire time.

[00:39:37.390] – Kriste Goad

That’s a lot of energy.

[00:39:38.920] – Todd O’Neill

Hundreds of thousands of gallons of water, which goes in cold and comes out warm. And then all the energy. Enough energy. I saw one to do some of the ChatGPT. Enough energy to power the country of Argentina for a month. I mean, there’s a lot of elements here.

[00:40:02.430] – Kriste Goad

Speaking of different elements, there’s something know, like ethical use and the conversations around regulations and everything around intellectual property. And you said something to me as we were prepping for this podcast around intellectual property. OpenAI ChatGPT. It’s not seeing intellectual property, right?

[00:40:28.310] – Todd O’Neill

No, it’s something that we. I think it’s important for people to remember that intellectual property paintings, anything that’s copyrighted, any information, if it’s on the Internet, it’s been digitized. Right. The stuff that you don’t have to worry about being in, like an image generator, is that painting that you have hanging on the wall that no one’s ever taken a photograph of. Because today, if you take a photograph, it’s digital pretty much, right? It’s digits and you’re probably going to share it with your family over the Internet and it will eventually get collected. So all of this intellectual property is. Yes, it’s certainly objects, paintings and books and other things, but ultimately it’s data. And that’s what OpenAI is thinking of. It’s data to feed their model. They’re not thinking. They didn’t go out and say, give me 10,000 pictures. No, they were like, we need data from 10,000 images because we can convert that data into a representation of a different image because that’s what the large language model is doing.

[00:41:57.340] – Kriste Goad

And that’s what we’ve been entering all that data in so that our products and our brands and our clients can be found. Correct. Our content can be found through, like to make it all tags and SEO and alt tags on images and all the things is now what’s powering this? Right?

[00:42:16.330] – Todd O’Neill

Yeah. And what I mentioned to you, I think when we’re getting prepped is what people need to remember, is that OpenAI, I’ll just use them as an example. But this is for all the companies that went out and created these models. They crawled the Internet with a different kind of crawler than Google uses. But they basically crawled the Internet. Right. They didn’t tell anybody they were doing it. They didn’t ask permission of anybody to do it. They didn’t ask permission of the owners, the copyright owners, they give no credit to any of the copyright owners. Even attribution, they’re not compensating any of the copyright owners and their argument. Most recently, Sam Altman from OpenAI said, if you don’t let us do this, if you don’t let us use copyrighted information, then our industry will not be successful.

[00:43:26.770] – Kriste Goad

You should earn all your money on the backs of everybody else’s hard work.

[00:43:34.950] – Todd O’Neill

Where does that happen? I mean, if you were in school and you copied somebody’s paper and turned it in as your own work, you would fail.

[00:43:43.910] – Kriste Goad

Would fail and get kicked out.

[00:43:46.180] – Todd O’Neill

And get kicked out, yeah, but that’s not happening here.

[00:43:49.720] – Kriste Goad

Okay, where’s all this going to go, do you think? I know. You got to track it. One of the questions I have for you is like, what are you reading? And how do you stay on top of all this?

[00:44:05.230] – Todd O’Neill

Wow. Yeah, my head hurts. I talked to a colleague about this. We’re teaching in the program and we meet every Monday to talk about our classes and our students and then everything that’s going on and basically our brains hurt. We read newsletters, countless newsletters. Every morning I have to go through my email inbox, not necessarily for messages because I get maybe 10%, maybe 20% of my inboxes are actual messages.

[00:44:43.290] – Kriste Goad

Right.

[00:44:43.880] – Todd O’Neill

Rest is clogged with, unfortunately, ads, spam junk. But a big chunk of it is newsletters on current trends, tech trends and that sort of thing. And some of the ones I read, a guy named Gary Marcus. He does one called Marcus on AI. It’s on substack. It’s very good. He actually did the research with another gentleman that was just released. It was through the IEEE, which is like the electric engineering organization. It was more like a white paper about image generators and how uncannily accurate they may be. They were able to reproduce almost exactly stills, like photographic stills, behind the scenes stills from feature films.

[00:45:39.210] – Kriste Goad

Crazy.

[00:45:40.140] – Todd O’Neill

Yeah, pretty nuts. Another guy I read a lot is a guy named Christopher Penn. He’s got something called the almost timely newsletter. He used to be very much, he still is a marketing kind of guy. Did a lot with SEO and analytics and very online focused. But he’s doing a lot of stuff around AI now and he’s got really excellent stuff on writing good prompts and what’s going on in the industry and what is ethical and what’s not and what’s the real business behind all this. He’s really quite good. That comes out once a week. Gary Marcus is publishing two or three times a week.

[00:46:24.510] – Kriste Goad

Oh, wow.

[00:46:25.290] – Todd O’Neill

And it’s extensive. I mean, they’re not like 500 word blog posts. They’re like in depth stuff. And then another guy, two other guys, Charlie Johnson does another substac called Untangled and he actually has an ebook called AI Untangled, which kind of does what the newsletter says he’s untangling. He’s like unpacking some of the big issues and what’s going on. And then Casey Newton has the platformer. It’s not quite a newsletter, it’s more like a blog. It’s quite good. But listen to a lot of podcasts. One that has Casey Newton on it called Hard Fork, the New York Times with a guy named Kevin Roose. Kevin Roose made a lot of news last January where he started working with Microsoft’s AI tool that was built into Bing and got it to admit that the AI called itself Sydney. And Sydney tried to talk him into leaving his wife. No, totally serious. He told him, you don’t really love her.

[00:47:35.890] – Kriste Goad

It’s like that Joaquin Phoenix movie coming true.

[00:47:40.130] – Todd O’Neill

Yeah. Oh, yeah.

[00:47:40.930] – Kriste Goad

No, it was Her, wasn’t it?

[00:47:44.850] – Todd O’Neill

Yeah, it was creepy. Very creepy. Another one that I like. It’s more commentary, but they do news. Kara swisher and Scott Galloway, they do pivot. That’s from New York magazine. That’s always good, always covering hot topics about the tech world, but a lot about AI. And then I’ve been listening, started listening to this. These are shorter things. Those other two are about an hour long. But this last podcast, what next? TBD from Slate is shorter, a little more bite size, 20 minutes or so. And they cover some good topics around AI, but also other kind of tech things because AI is really creeping into all different elements of things. So you kind of have to keep an eye on all of it. I haven’t really seen anybody doing something that is like AI for communication professionals. I haven’t really found any newsletter that’s doing that specifically, but all of them are talking about it. There’s a content marketing thing by Robert Rose, content Marketing Institute. They talk about AI as part of content marketing. Social media examiner talks about AI as part of social media. The whole idea of AI being used for search is scaring the bejesus out of SEO people, because if you can just ask an AI for something and it’s accurate and it puts it into paragraph form, why do you need a search result page?

[00:49:28.150] – Kriste Goad

So I was having the same conversation at an entrepreneur’s dinner about a year ago, and the woman sitting next to me, she has an online publication. And so she had been experimenting right after this came out, she was spending her weekends experimenting. And that was one of the ways she was experimenting because her online publication, she needed the search to be good and the SEO to be good, and she was finding it very effective in that department.

[00:49:59.160] – Todd O’Neill

Yeah, it’ll give you full. Why click a link to get to an article? If you can just have something like a ChatGPT or tool built into Bing or whatever, that will just give you an article. Done. But why go to a website?

[00:50:21.510] – Kriste Goad

We’re going to need a whole nother podcast to talk about this.

[00:50:24.810] – Todd O’Neill

Oh, it’s crazy. I was teaching SEO last spring. Yeah, ChatGPT hit hard. This is around, we were only about three or four weeks in and I said, listen, guys, you all have been playing around with generative AI. And they’re like, yeah, a little bit, not so much because the academic community at the time was freaking out.

[00:50:45.950] – Kriste Goad

Yeah.

[00:50:46.860] – Todd O’Neill

Oh, people are going to write their article. It’s like, stop, just stop. And then the tool people, we have a tool in our learning management system that detects plagiarism by looking at a database of articles that have been written and doing text matching. Well, it came out in March and said, oh, yeah, we got this AI covered. We can totally detect it and pick it out. Like, no, you can’t. No one can.

[00:51:19.910] – Kriste Goad

Stop.

[00:51:20.380] – Todd O’Neill

Just don’t even say this. But in February, a few weeks into my SEO course, I just said, so we’re going to continue on with our curriculum this semester, but we will chat about, we’ll talk about AI. But just so you know, I have to rewrite this curriculum for next spring because AI is going to change search engine optimization. It is. Google just integrated AI into, I can’t remember the name of their tool into their search engine. You can use a ChatGPT to come up with something, and then it allows you to press a button to say, search for this information or verify it and to bring this stuff in.

[00:52:09.290] – Kriste Goad

So have you rewritten that curriculum yet?

[00:52:12.040] – Todd O’Neill

I’m working on it. Yeah, I’m delivering it. Yeah. I’ve gone through and touched on things, but a lot of it, frankly, right now, the tools aren’t quite good enough yet to be a competitor for search engine optimization because they’re just even as good as they are. They’re not quite that good yet. Remember that one month after ChatGPT was released, it had 100 million users?

[00:52:47.450] – Kriste Goad

Insane.

[00:52:48.190] – Todd O’Neill

It took Netflix five years to get 100 million users. Yeah.

[00:52:54.230] – Kriste Goad

And the model is learning. It’s operating faster. I’ve been reading about that. Yeah.

[00:53:00.490] – Todd O’Neill

There’s this concept called Moore’s law, where every 18 to 24 months, it has to do with electronics, really? But every 18 to 24 months, the power of our phone will have doubled. Right.

[00:53:14.800] – Kriste Goad

Right.

[00:53:15.260] – Todd O’Neill

The resolution of our pictures will be twice as much. AI is not on that track. It’s every month. It’s continually learning.

[00:53:28.770] – Kriste Goad

I’m getting more and more anxious the longer we talk here.

[00:53:31.970] – Todd O’Neill

But see, that’s the cool thing that, wow, this can really learn. This could really do something really cool. But we have to go back to, it’s all our data, and this country has no record of managing or regulating data in any way. We don’t do it, and it learns really fast. And these are commercial products based on what the companies say are experiments, and they’re charging us for. And, like, Microsoft does this all the time. They release the beta version of their product, see if we could break it. And when we do, they fix it.

[00:54:21.120] – Kriste Goad

Right.

[00:54:22.710] – Todd O’Neill

They’re constantly fixing it. And the way it works, it basically is constantly accelerating.

[00:54:33.350] – Kriste Goad

Right.

[00:54:34.140] – Todd O’Neill

It’s not on a steady path. It’s constantly. Constantly.

[00:54:37.840] – Kriste Goad

Like a rocket ship.

[00:54:39.180] – Todd O’Neill

Yes. And it’s really scary. There’s really no stopping it. The only way to stop it is for the government to step in and tell OpenAI. Great. Now drag your large language model into that trash icon and get rid of it.

[00:54:59.220] – Kriste Goad

Yeah.

[00:54:59.940] – Todd O’Neill

Now you can start having a product and you’re going to have to rebuild it and spend the billions of dollars you spent already in development and rebuild that large language model in an ethical way where you give compensation to the people who are building it. That’s not going to happen.

[00:55:18.890] – Kriste Goad

This will be fascinating to watch this unfold.

[00:55:22.050] – Todd O’Neill

Yeah. Sorry. We have been going on for a.

[00:55:24.260] – Kriste Goad

Long, I know, this has been so amazing, Todd. Yeah, we’ll shut it down. Just promise that you’ll come back and we’ll continue the conversation.

[00:55:32.910] – Todd O’Neill

Yeah, no, I love talking about communication. Communication is my life. That’s what I’ve done forever and ever. And I love it and I love teaching it. And I come to work every day happy.

[00:55:50.670] – Kriste Goad

That’s amazing. They’re lucky to have you, and your students are lucky to have you.

[00:55:56.020] – Todd O’Neill

Yeah, it’s a good time.

[00:55:57.570] – Kriste Goad

So I really appreciate you spending your time with me today. And you’ve given us a lot of great information and newsletters and things to take a look at and read and listen to. We’ll include those for our listeners in our show notes, and we’ll be sure to include links on how to get in touch with you and all your information that you want to share. Good that you have given us permission to share. But thank you so much, everybody. Just be sure to check out our website, growwithfuoco.com/howitsdonepodcast or wherever you get your podcast. So as I mentioned, be sure to follow the podcast. Be sure to check out the page, be sure to check out our show notes. There’s going to be lots of great links to lots of great information that we’ve covered here today. Again, my guest has been Todd O’Neill, associate professor of Interactive Media and coordinator of the Interactive Media degree program for the College of Media and Entertainment at Middle Tennessee State University. Todd, any final parting words?

[00:57:09.030] – Todd O’Neill

Just embrace the future. You know, just put your arms right around it and love it because it’s coming and just, just go for it.

[00:57:21.260] – Kriste Goad

I love it. Thank you so much for that. And thank you, Todd.

[00:57:24.770] – Todd O’Neill

Yep, thanks, Kriste.

[00:57:26.450] – Kriste Goad

That’s it for now. Thanks so much for listening. We’re looking forward to keeping great conversations coming your way as we grow this podcast. There’s even more great content from our conversations on our blog. Be sure to check it out at growwitthfuoco.com. That’s growwithfuoco.com. Stay tuned until next time. And no matter what, stay curious.

• • Blogs and Podcasts
    • Marcus on AI
    • Generative AI has a Visual Plagiarism Problem
    • The Almost Timely Newsletter
    • AI Untangled
    • The Platformer
    • Hard Fork
    • Kara Swisher and Scott galloway: Pivot
    • What Next: TBD
    • Content Marketing Institute
    • Social Media Examinar
    • AI Guidance from Healthcare Communications Association
    • PRSA AI Ethics and Guidance 
  • Find Todd on:
    • LinkedIn
    • http://toddoneill.com
  • Find Kriste on:
    • LinkedIn
    • Twitter
  • Email Kriste: kg@growwithfuoco.com

(03:07) Todd describes the importance of being an early adopter of technology along with the role it plays in how he teaches his students

(05:50) Kriste and Todd discuss why most college grads aren’t up to speed with emerging technologies

(07:26) Todd and Kriste talk about how it can be difficult to convince clients to use emerging technologies, relating it to the early days of social media

(14:49) Todd tells Kriste why it is important to separate the technology (Large language models and AI) from the product (ChatGPT, DALL-E, Claude)

(18:55) Todd shares examples of how AI tools can potentially reduce copyrighted material and best practices for marketers and PR professionals using the tool

(21:19) Todd shares a real life example of ChatGPT’s impact in colleges 

(22:41) Todd describes the privacy implications of using AI products, comparing them to Facebook

(24:11) Todd discusses the future of AI in private label large language models customized for companies

(28:40) Todd and Kriste discuss the ethical considerations, including the importance of disclosure when AI is used for content creation

(30:35) Todd shares a real life example where a company outsourced content creation to another agency using AI, leading to a trail of misinformation 

(33:12) Todd and Kriste go over the PRSA AI usage guidelines and what the biggest takeaways are for marketers and PR professionals

(38:41) Todd tells Kriste about Middle Tennessee State University wanting to make their own large language model and the environmental consequences of that

(40:28) Todd tells Kriste about the unethicality of the way OpenAI’s ChatGPT was trained

(43:49) Todd and Kriste discuss the lack of regulations and the challenges in of regulating and managing data in the abscess of proper frameworks

(50:21) Todd tells Kriste how he has to rewrite his SEO syllabus due to the shake up of AI powered searches.

(54:49) Todd shares why we should embrace the future and the technological advancements