Ep 40: How to use AI in Marketing with Laura Gemmell

Powered by RedCircle

Shownotes

In this episode, Joyann is joined by Laura Gemmell and together they delve into the world of artificial intelligence.

They discuss the construction of chat GPT, offer practical tips on how to use it and explore a myriad of ways AI can be utilised. They also address the issue of bias in AI. 

You can watch the podcast here.

Sign up to our newsletter here.

You can find the transcript here.

You can find Laura here:

Laura’s LinkedIn

Laura’s website

Taught by Humans LinkedIn

Taught by Humans website

Taught By Humans YouTube

Useful links:

Chat GPT

Reinforcement learning

Drake and The Weekend Deep Fake

‘Be Right Back’ Black Mirror Episode

Claude

GPTs

‘White Christmas’ Black Mirror Episode

Supabase

Enniskillen Christmas Ad

You can find Joyann here.

Transcript

00:00:00:00 - 00:00:32:07

Joyann Boyce

Welcome and welcome back to the Marketing Made Inclusive podcast. I am your host, Joyann Boyce, and today we are joined by what I just realised is a first non-marketer, Laura, oh see I pause because I stumbled Gemmell. This is becoming an unfortunate tradition, but also it's because I overthink it. Laura Gemmell is a data scientist who is a friend of mine and I wanted to have her on the podcast so you can tell all things.

 

00:00:32:12 - 00:00:42:17

Joyann Boyce

I, I do think you are somewhat of a marketer, but I'll let you self-described if you feel that you are or not Laura, introduce yourself to the people.

 

00:00:42:19 - 00:01:11:12

Laura Gemmell

I'm Laura Gemmell, I'm the founder of Taught by Humans. I'm currently saying I'm the technical CEO, which is the chief everything officer because I literally do everything for the business. So, yes, marketing does fall under everything that a startup founder has to do. As Joyann mentioned, I am a scientist or a technologist by trade. I have a Ph.D. in educating the public about artificial intelligence, on the skills needed to work in these areas and Taught by Humans is an AI data confidence company.

 

00:01:11:12 - 00:01:18:23

Laura Gemmell

We are building a product to help people learn in a personalised way so they can become AI and data confident in their jobs.

 

00:01:19:00 - 00:01:55:17

Joyann Boyce

Absolutely love it. And who better to have on the podcast with me to discuss all the things AI because I know sometimes I am some of my marketing friends go to person for AI questions. Laura is my go to person. So my first question, Laura is the generative AI boom, right? We know it's been around for a while, but I feel like generative AI is the one that has been impacting marketing one of, the most, like, how do you see it and, and what are your vibes on it?

 

00:01:55:19 - 00:02:19:12

Laura Gemmell

Yeah so I think, what, when we're talking right now, we've had about a year of generative AI being very mainstream so it's been around for a while but we've had a year since the Chat, the first Chat GPT was released and everybody went quite intense about this whole thing. To me I would say I think why this boom has happened and why it's impacted things like marketing is because people can actually touch this one.

 

00:02:19:14 - 00:02:41:00

Laura Gemmell

And what I mean by that is Chat GPT was an interface where people can literally type in a way that they do other things on the internet and feel how this AI was talking back to them and things like that. Before we kind of had things like neural networks and which, which did great things in areas like predicting whether an image had a cancer tumour in it and things like that.

 

00:02:41:00 - 00:03:02:12

Laura Gemmell

But again, people didn't really get to touch them. They just got told this was the result. So the algorithm and it seemed a bit detached, so generative AI people could touch, they can do fun things with, they can make images, they can tell them jokes, they can get it to write for them. So to me it just feels a little bit more real, a little bit more every day than previous iterations of AI before.

 

00:03:02:14 - 00:03:32:08

Joyann Boyce

I think that's interesting because the touch element for me, I always say to people well we've used, as marketers, we use Google, we use Google Analytics. But I realised a lot of people didn't know it was involved in that. So what, what do you think? And it is funny because I sometimes think marketers ruin everything. But do you feel that because it's marketed as an AI, like literally has as AI Chat GPT, does it have AI in the name?

 

00:03:32:10 - 00:03:35:15

Joyann Boyce

It doesn’t. I just realised that. Sorry. I know,

 

00:03:35:15 - 00:03:52:00

Joyann Boyce

Because we know Google and we know SEO and we know all those things has AI integrated in it. What's the difference from using Google now to using Chat GPT like why are people now clocking the AI?

 

00:03:52:02 - 00:04:13:01

Laura Gemmell

I think it's the personality part and yes, I don't think AIs have personalities, but you feel like you're talking to something a little bit more human than when you're typing into a search bar. So people want to be friends with it. It's like when people talk to Alexa, a lot of people said please and thank you, even though there's absolutely no need to, it's because they feel like they're talking to a personified something.

 

00:04:13:03 - 00:04:31:00

Laura Gemmell

And I think that's the Chat GPT thing. Like when you, when you chat to service bots in the corner of the screen and you end up, well I always end up writing human capital letters because they annoy me so much. Yeah, that's how I get it to talk to a person. I just keep writing it in capital letters until it listens to me, whereas with Chat GPT.

 

00:04:31:02 - 00:04:49:19

Laura Gemmell

It does kind of, it corrects itself, it apologises, it does all these things that make it feel a little bit more human, which make people either trust it more or be more wary of it, depending on how you feel about it. But yeah, when you type into a Google search bar or when you use a bit of software, it's very removed, right?

 

00:04:49:19 - 00:05:03:23

Laura Gemmell

You don’t see what's going on. It talks back to you. It’s nice to you. It's like, here's your results. You've done this wrong. And I think it's just that bit of it that makes Chat CPT and all the other assistants feel a little bit different.

 

00:05:04:00 - 00:05:30:20

Joyann Boyce

Because, yeah, with Google you just get delivered the results and we as marketers are always trying to rank, but there's no way for us to gamify or get into Chat CPT to get our content to rank. So high level. And Lord knows sometimes I'm asking questions more high level than we normally would have conversations, but so they, to create Chat GPT.

 

00:05:30:22 - 00:05:49:22

Joyann Boyce

To my understanding. They scraped the internet and that means that they went and they got data from everybody's websites and chucked it into a machine. And then the machine learned from that and now answers all our questions based on the Internet. Is that a way you could say it worked? How? How?

 

00:05:49:24 - 00:06:07:18

Laura Gemmell

Yeah, I think there's a few interesting stats in the middle that I can talk about that quite high level that I find really interesting. So whether the entire internet was scrapped or not is a contentious point, right? Because it's said all the time. Open AI constantly say that they did not scrape the whole internet, but I don't know.

 

00:06:07:18 - 00:06:32:19

Laura Gemmell

The answers are very Reddit leaning, aren't they? So like the website Reddit? The answers very much answer like that. Or if you're trying to be friends with it, it really talks to you like the way bloggers write. And so I don't know. But one of the interesting things is they built this algorithm and the algorithms are called GPT. So Chat is actually the chat bot interface that they've built for it to be friendly and how they got it to behave like this is they built another algorithm.

 

00:06:32:21 - 00:06:58:22

Laura Gemmell

So there was a point when humans were telling it what to do and that's called reinforcement learning. So you're saying right. That's wrong. That's right. That's right. That's right. That's nearly right. And they realised they were doing it in such a way that they could build another algorithm that was almost like the checking algorithm. So they built another algorithm that then could train the GPT algorithm to be better, which I think is really interesting and probably why it got so supercharged so fast.

 

00:06:58:24 - 00:07:18:24

Laura Gemmell

But to me that was very interesting. So that's a reinforcement learning using multi-bots, I think is the term these days. If you want the technical term. But yeah so they built all these different systems that can actually make their algorithms better almost automatically after they've built the algorithms. So a few extra steps in the middle, not just a script.

 

00:07:19:00 - 00:07:23:09

Laura Gemmell

The Internet made, made it into a good algorithm, but yeah, very interesting.

 

00:07:23:11 - 00:07:48:15

Joyann Boyce

It's the bot training it because it’s funny thing like, and I say this jokingly, but I sometimes blame marketers for the reason that the algorithms are biased in any particular way or have a particular tone of voice because I'm like it learned of, of our content is all the times we were trying to write blogs in a friendly tone of voice and use extra words that were not necessary. I don't know if that's our fault, actually, we use less words.

 

00:07:48:17 - 00:07:52:18

Joyann Boyce

It’s probably researchers. I'm going to tend to research content.

 

00:07:52:20 - 00:08:24:08

Laura Gemmell

Also, I think they set settings so that the chat GPT version of the algorithms is actually quite long winded and you can tell it to not be so long winded. There are ways you can set that, yeah, I hate how long it talks. I hate how many words it uses. It's so annoying. But yeah, I think that was just a version of it and I think the newer version is slightly shorter because they've realised that people have, people don't want to read reems and paragraphs and everything doesn't need to be a bullet pointed list.

 

00:08:24:10 - 00:08:27:00

Laura Gemmell

Like it seems to love to give out.

 

00:08:27:02 - 00:08:46:07

Joyann Boyce

Its, its whole life story. At the time of recording this, it’s the 27th of November, the speed of which AI is moved is very quick, I just wanted to put that in there. But you mentioned that you got it to not be as wordy. How have you been using it? Because I know you used it for one particular experiment.

 

00:08:46:07 - 00:08:53:07

Joyann Boyce

I've seen on LinkedIn where you get it to write LinkedIn posts as you. What uses and what tips do you have?

 

00:08:53:09 - 00:09:24:02

Laura Gemmell

So if you're using Chat GPT as the interface we can talk. My tips would be you can click on your icon in the bottom left and you can set some of the system prompts it's called. So those are the things it believes. So mine is I live in the UK speaking British, English because if you use it a lot, you know it always ads zs and spells things American and if you're using it to make copy and stuff, it's really annoying to have to go and copy and paste and edit all those ads and put everything into British English.

 

00:09:24:04 - 00:09:42:24

Laura Gemmell

I've also told it to assume I know nothing. So it's always explaining things and I've told it to write a short answers as it can to get the, to get its point across. You can play around with these. You can have it so that you set up as a marketing assistant so that it always talks in marketing speak. You can have it so that it talks like a CEO or a scientist.

 

00:09:43:01 - 00:10:03:14

Laura Gemmell

But I think that's really interesting to play around. And you can keep the settings set all the time and then you don't need to include them whenever you're talking to it, which is called prompting. So whenever you talk to Chat GPT, we call that prompting. So for example, I consider myself a prompt engineer. I am joking. So this is meant to be said as a joke.

 

00:10:03:16 - 00:10:05:02

Laura Gemmell

Just so we're clear.

 

00:10:05:02 - 00:10:06:15

Joyann Boyce

Prompt engineers are things.

 

00:10:06:17 - 00:10:23:15

Laura Gemmell

They are a thing. So I've been playing around quite a lot with how to get it to do what I want. So and as you Joyann already mentioned, I have a LinkedIn post. It does every Monday, but I also wrote a blog series which basically started with me saying that I thought it was stupid and why would I use it instead of Google?

 

00:10:23:17 - 00:10:40:19

Laura Gemmell

I think on my own like episode or blog 12 and I've gone through the most of the motions of actually starting to enjoy it. A lot of people open it up and they type something in that you can type into Google, right? I want a list of this top this, so my mum is like, what's the best show on Netflix?

 

00:10:40:21 - 00:10:56:23

Laura Gemmell

That's actually a waste of time for me to ask Chat GPT, right? It's not useful. It has the same results I would get on Netflix or on Google, sorry, but you can personalise it, you can tell it things about you, you can tell it what else you've watched, things like that. But what I find really useful is if I wanted to do something.

 

00:10:56:23 - 00:11:16:21

Laura Gemmell

So one of the things I got it to do recently was to, oh I’ve totally forgot. Oh, to write a transcript from the transcript to make subtitles so that I could put them on a YouTube video. So the first thing I did was ask it, how do I use you to do this? So how do I use you to make subtitles for a YouTube video?

 

00:11:16:23 - 00:11:36:11

Laura Gemmell

And then it gives you a step by step guide. This is what I need to do that you need to tell me these things. And that's really helpful because then you can actually send it back. So I always ask it how to do it. And then when I write the prompt, I write it like you, I want this and then a clarification sentence and then a few things maybe that I want it to do.

 

00:11:36:11 - 00:11:54:07

Laura Gemmell

So maybe I don't want it, usually I give it and I don't want you to do this. So what you think it's going to do is going to annoy you. So if I'm feeding it in, for example, write in samples, I can say, I don't want you to respond until I say I am finished and things like that and then give it some tips so one could be explained all the jargon.

 

00:11:54:07 - 00:12:14:23

Laura Gemmell

Assume I'm a software engineer and I know all of this, assume I'm a marketer and I know common marketing terms. So my, my prompts actually end up being about, the sacrament always ends up being about a paragraph, and it can sometimes have an additional thing attached to it, like some code or the transcript or a blog, but that took me quite a while to refine, to get it to do what I want.

 

00:12:15:04 - 00:12:35:19

Laura Gemmell

And I also have to proofread everything I send in to check that I haven't done it in a Northern Irish way. So as you can hear from my accent I’m Northern Irish and we seem to use words incorrectly and also use verbs slightly incorrectly, I never notice it until I write it down, but GPT can't deal with it for a little bit and gets really confused and then makes up what I meant.

 

00:12:35:21 - 00:12:58:09

Laura Gemmell

And another useful thing to tell is don't, don't make things up. So GPT 3.5 has a tendency to fill in the blank. So it was trained to always give an answer. GPT four and onwards isn't as bad, but there's a term hallucinations, which is when an, when one of these generative AI just makes things up that isn't there and it does happen quite a lot actually.

 

00:12:58:15 - 00:13:19:14

Laura Gemmell

We've been training an algorithm to do personalisation of education for a very long time. If we didn't have the content, someone asked for it made up YouTube links. YouTube videos and YouTube links. So if you're using GPT 3.5, which is the one on the free chat CBT version, make sure to tell it not to lie or not to make things up so that you don't get incorrect information.

 

00:13:19:16 - 00:13:40:15

Joyann Boyce

The not to lie bit is so help, so like, the amount of times and especially when it gives you a lengthy answer and you're just like all of this is B.S. I did not need any of that. But something you said I, I think I'm still stuck in my bad, not bad habits, old habits, where I still go to Google to find out how to get the prompt to do the thing.

 

00:13:40:19 - 00:13:44:19

Joyann Boyce

I don't ask it, so I’m definitely going to add that step in.

 

00:13:44:21 - 00:14:01:03

Laura Gemmell

So for my LinkedIn post that I do every Monday where I get Chat GPT to write me a mundane, efficient LinkedIn post like me, I got it to write the prompt. I said, how would I, how, what prompt do I need to give you to do this? So I copy and paste that every week and edit some bits about it.

 

00:14:01:03 - 00:14:19:09

Laura Gemmell

But now we have an ongoing Monday conversation where I give it feedback. on, it's a post from last week and it's still very biased. And like I told, I was Irish one week because people told me the post sounded like I was a Californian surfer dude. So I was like, I'm Irish, do it again and it included a shamrock and stuff in it.

 

00:14:19:09 - 00:14:26:20

Laura Gemmell

So I was like, cool, calm down, I can, I can be Irish, but it doesn't need to spill through every sentence of the post.

 

00:14:26:22 - 00:14:48:14

Joyann Boyce

You've landed on my next question. So the bias that's within the models and within the AIs, if there was something another bot training that bot, where did the bias come from? Is it from the content or is it because the bot doesn't understand, how did it get in there?

 

00:14:48:16 - 00:15:13:03

Laura Gemmell

Like and I will say like I'm not an expert on how exactly this algorithm is trained and open AI obviously, because it's commercially sensitive, quite, quite, keeping it a secret. The internet is not a place of non-bias. I don't think there was any point in the in the training of the model where they went through and said this is biased, this is biased, this is biased and told it not to do it, not to include things.

 

00:15:13:05 - 00:15:39:11

Laura Gemmell

Interestingly, though, they ran a survey there, they ran some research like an independent university on Chat GPT. And it's apparently very left leaning politics wise, which is, which is interesting. And yeah, so it seems that they scraped certain things from the internet. So there's definitely bias added in from the data itself, right, because if they just scraped the Internet or if they're not willing to share where the data came from, then we have to assume that is biased.

 

00:15:39:11 - 00:16:02:06

Laura Gemmell

If there's been no checking that it wasn't biased because one world is kind of biased. And two, the Internet is very, very biased. And the training algorithm was made to, made to copy a human behaviour. So again, that most definitely could have been bias and put into that that aim was just to make the algorithm more accurate, as far as I'm aware, which again, doesn't account for any bias.

 

00:16:02:08 - 00:16:29:12

Laura Gemmell

It is a very complicated one because I guess obviously bias isn't a good thing, but if you want a Chatbot to be realistic about the Internet, then it probably is going to be a bit biased. I'm not one of those people that plays with Chat GPT and asks it bad questions, so I don't know if my bias is and I guess some things that it does pick up on is it assumed I was a man because of my posts on technology and yes the Irish one I've already hit on.

 

00:16:29:14 - 00:16:51:19

Laura Gemmell

And yeah, if you ask it to write different text for different, different marketing social media platforms, it does kind of assume the general user and get like that. So for Instagram it assumes a female. And so it's interesting that it does that, but yeah I haven’t played off, I know loads of people who included Joyann, who tried to make it be biased, but I'm not one of those people that has done that.

 

00:16:51:21 - 00:17:19:15

Joyann Boyce

Oh, I have. I have seen things that GPT 4 and 3.5 have said. I think a good one for marketers to test the bias, and that is ask it to make a marketing persona based on high level demographics. And it is very, very interesting. The worst ones that I've seen, as I said, to make a marketing persona on a deaf Black woman.

 

00:17:19:17 - 00:18:04:01

Joyann Boyce

And yeah, it, it said that she transcended sound and skin. And I was like I don’t, there’s so many layers of things are not logical in that. The latest one we experimented was describe a Black Harry Potter in 100 words. No that one was. Yeah, yeah. But the I think it's really good to note that Open AI hasn't been transparent and that's like super frustrating because at least we could know going into it, using it for content creation that, you know, okay, if I'm going to write a blog on Apple Computers, it might go and replicate content.

 

00:18:04:01 - 00:18:23:20

Joyann Boyce

By MQBHC it was a YouTube reviewer about Apple and stuff like that. And it makes it very tricky for a marketer to know anyway to trust that the content they're getting out of there is 100%. Are there any way, other ways you can think of using it for marketing or any other ways you've used it for marketing?

 

00:18:23:20 - 00:18:33:24

Joyann Boyce

In your case you mentioned LinkedIn post, blog, transcripts of YouTube videos, which I didn't know that one it could do.

 

00:18:34:01 - 00:18:50:09

Laura Gemmell

So I would just call it, I only use it for one LinkedIn post and I make it very clear that Chat GPT actually wrote that LinkedIn post because it doesn't really sound like me despite me trying and I think you're touching on something that we shouldn't really get it to do the creative side of any job, particularly in marketing.

 

00:18:50:09 - 00:19:09:19

Laura Gemmell

So I don't get it to write blogs for me. I sometimes get it to read blogs for me and proofread them for me. So I'm dyslexic and I'm absolutely terrible, like getting things right the first time. I quite often use the wrong word, or as I said, I tend to use weird colloquial isms that shouldn't be used and, and things like that.

 

00:19:09:19 - 00:19:29:17

Laura Gemmell

And I have a very lovely husband who mostly checks my stuff for me. But Chat GPT is a, is a great second whenever I need someone to help. So checking anything and sometimes getting, writing your blog post that you're going to edit because staring at a blank screen is sometimes the bit that is the blocker for a lot of people.

 

00:19:29:19 - 00:19:53:18

Laura Gemmell

So asking it to write a blog post and then being like, oh my God, this is terrible and then editing it yourself, but having some semblance of a structure that probably makes sense because it's, it's not terrible structure. It's terrible at like flowy, nice language in the middle. Also it's quite good for SEO, although I don't think because it's, because if you think about how it was done, it's very logical and algorithmic.

 

00:19:53:18 - 00:20:17:12

Laura Gemmell

Anyway, one thing to know is that a lot of websites are, but particularly website builders like Wix and things are blocking anything. So my website's hosted on Wix at the minute and I cannot use Chat GPT to do the SEO on it, because it's blocked. Yeah. So a lot of websites are blocking web scraping and stuff like that which is how this type of SEO thing, would generally be done. It’s quite good at that.

 

00:20:17:14 - 00:20:34:19

Laura Gemmell

If, if it can get on your website, it's also quite good at doing a website audit like saying like, oh, it needs to have one hitch, one oh, this bit doesn't make sense. You might want to add a tagline higher up the page and things like that. Planning, I quite like it for planning again, even if it's just to give you a plan.

 

00:20:34:20 - 00:20:52:24

Laura Gemmell

You're like, that is terrible, but why would I do that? At least you've got something to start with. And good for competitor research actually. So now that it can access the Internet, you can actually talk about competitors, you can ask it to try and find out colours and websites and stuff and things like that, which I find really helpful.

 

00:20:53:01 - 00:21:09:19

Laura Gemmell

Basically, anything repetitive or anything that's going to take you a long time. The other day I had like a list of emails that I had signed up at an event and they'd been sitting on my desk for a month because there was 40 to 44 emails on it and I could not be bothered typing 44 emails into my computer.

 

00:21:09:21 - 00:21:25:06

Laura Gemmell

So I was trying to work out a way, I could do it quicker. And taking a picture on Google Translate on my phone and getting it to output some data and then I copied and pasted the data into Chat CBT and told it what it was. These are supposed to be emails that were signed up in an event.

 

00:21:25:06 - 00:21:44:04

Laura Gemmell

This is the output from Google Translate. Can you actually work out what the emails are supposed to be? I did a pretty good job. I think there was only ten that it got wrong and mostly they were from handwriting. Were Google's fault, not Chat GBT’s fault. So something that I think would have taken me an hour and a half or something ended up taking me 5 minutes.

 

00:21:44:06 - 00:21:47:07

Joyann Boyce

So it’s time saving, not time replacement.

 

00:21:47:09 - 00:21:48:05

Laura Gemmell

Yeah.

 

00:21:48:07 - 00:22:12:16

Joyann Boyce

Something you mentioned that I'm curious on. So if and I think a lot of this is really good for like the all in one marketer, the one who's doing 152 roles and they could just get the top level stuff. But I'm curious for the marketers that are in large corporations where there's a lot of protection on copyright and so forth, how do you see that relationship work?

 

00:22:12:16 - 00:22:25:23

Joyann Boyce

And I know it hasn't been defined yet, but if someone wants to put a piece of content that they've written into chat GPT is it still theirs? Is, what your thoughts? I know, I know it isn’t law yet.

 

00:22:26:00 - 00:22:46:03

Laura Gemmell

Yeah, I think so. I think like it's, it's going to go on the internet anyway, right. So when they put it on the internet it's still there’s right, isn't it right? And I think I'm very into I, I think copyright is very important and I did have a chat with a barrister about this really randomly a few weeks ago.

 

00:22:46:04 - 00:23:17:15

Laura Gemmell

He's really interested in all of this and he said that and it will always be very difficult to do for anyone to take any copyright cases against these algorithms because copyright law in the UK at the minute is very, very strict that it must be a clearly defined copied path and it would be really difficult to ever, unless there's like word for word copying, you can't really copy tone and style because that's actually not something you can like write down and protect.

 

00:23:17:17 - 00:23:38:00

Laura Gemmell

So he saw it as like unless it was literally taking paragraphs out of a book or something, it wasn’t going to be seen as like copyright. And if it's on the internet anyway, and anyone else can do that as well. So I think people still own the copyright. I think if you put it on the Internet anyway, it's probably fine to put it into Chat GPT.

 

00:23:38:02 - 00:24:03:23

Laura Gemmell

Just my personal opinion. I think some of the court cases going on at the minute are very interesting. So this is slightly going off marketing, but the, the fact that and in in the music world, the deep fake, the Drake and The Weeknd’s and that's been submitted to the Grammys and stuff like that. So how all of those things play out because Drake and The Weeknd didn't say they could be deep faked to deep fake this whenever you like.

 

00:24:03:23 - 00:24:24:12

Laura Gemmell

Copy someone's image and voice. So this TikToker, going to show that I'm not into, into the, into social media, this TikTokker whose name I can't remember anymore and has a deep fake account where he deep fakes music videos and this one went up and then got taken down like immediately because I assumed that the label were like threatening all kinds of things.

 

00:24:24:17 - 00:24:41:07

Laura Gemmell

But then the TickTokker submitted it to, to the Grammys to see if it would get through. And I'm not sure what the outcome of that is, but I think things like that are very important at the minute because that's going to set the tone right. Do we, because with the number of views that got on YouTube and stuff, it's probably one of the biggest songs of the year.

 

00:24:41:09 - 00:25:04:18

Laura Gemmell

But like can that be submitted to the Grammys? Like can it be, does Drake win an award if, if he was deep faked? Like how does that end up working? So those conversations are really interesting. And same with the writers strike in America. So not copyright, it's more like job protection a little bit. And so all of these things going on at the minute are going to influence copyright laws, they’re going to influence everything that happens.

 

00:25:04:18 - 00:25:25:22

Laura Gemmell

So the, how these things turn out is very important. So the writers are striking to protect themselves against being replaced from AI generated script writing. They don't want to end up just being the person who checks the AI generated script at the end. They want to be involved in the creative process and like completely agree. Have you seen things like creative by Chat GBT?

 

00:25:25:22 - 00:25:50:19

Laura Gemmell

There's no creativity. They miss like nuances, like humanness, like things that would be funny that it wouldn't quite understand. So there’s an excellent Black Mirror episode on this, and there's an episode where after someone dies, you can bring them back to life. And the episode starts off with like, I can't remember what, it’s like Jamiroquai or Bee Gees, like Guilty Pleasure song comes on in the car and the guy's like, oh, I love this song.

 

00:25:50:19 - 00:26:08:06

Laura Gemmell

And his wife's like, that's not really like in your case. And he's like, oh, guilty pleasure and turns it right up. And then whenever they fake generate him back in the future, the song comes on the radio and he says he hates it because you can’t actually get digging because by all algorithmic written down parts of it, he should hate the song.

 

00:26:08:06 - 00:26:27:23

Laura Gemmell

It's really against his personality to like the song, but he loves the song. So I think that's what's going to happen is if we keep using AI to do all the creative stuff and in marketing as well, because sometimes marketing that really touches your soul or it's really moving or it's really funny, I think it's going to end up missing all of those bits and it's just going to be like, oh, I'll go and buy that.

 

00:26:28:04 - 00:26:34:13

Laura Gemmell

But you've no emotional connection to why you're going to buy that or sign up for that or whatever.

 

00:26:34:15 - 00:27:05:07

Joyann Boyce

That's, that's so needed. A theory that I have and I was speaking to someone on the podcast about it was that marketing is going to go back to the psychology of it. It's going to go back to the core elements. But I'm curious to know if you could, because we're recording this and end of November and it's going to come out in February, if you predict, because I think I'm comparing AI years to dog years because I'm obsessed with dogs at the moment.

 

00:27:05:09 - 00:27:23:03

Joyann Boyce

If you can't predict what the next couple of months are going to look like in the AI world and then correlate that with how you think it might impact creatives, what do you think? Where do you think we'll be?

 

00:27:23:05 - 00:27:49:17

Laura Gemmell

So speaking very much on what's happened recently. So recap if people are listening to this in February and don't know, it's been a lot of things going on with Open AI. Who are the company who created Chat GPT, a lot of moving around the CEOs and then reinstating the CEO and the whole new board being formed. I think what that's actually opening up is space for other players, because a little bit of faith has been shaken in them.

 

00:27:49:19 - 00:28:16:19

Laura Gemmell

They still have a stronghold, but for example, Anthropic who make Claude is a chatty competitor, are currently doubling buying and lowering prices and making it easier for people to access. And they're big focus is AI safety, which isn't quite bias, but it touches on things, it touches on thinking about bias, I think is how I word it, so I see like things potentially moving towards it being a more equal playing field.

 

00:28:16:19 - 00:28:38:17

Laura Gemmell

Not everybody talking about Chat GPT. I also from a personal experience on talking to other people. We are going to start seeing bots that have jobs and what I mean by that is potentially within a marketing flow, there might be the writer, the editor, the something, something, the creative one and they're all going to have little personalities.

 

00:28:38:19 - 00:29:14:06

Laura Gemmell

I say this kind of facetiously because like they're going to have personalities given to them by a person and they're going to do specific jobs as opposed to trying to get Chat GPT to do everything for you. But so Open AI has also just launched GPTs. GPTs still uses Chat GPT, but I think we're going to get closer and closer to people being able to edit things more and make them work in a really specific way for them and the barriers to entry to that, which was originally you must be able to code, is getting lower and lower and lower as these companies are realising that's what people want.

 

00:29:14:08 - 00:29:33:01

Laura Gemmell

So in terms of creativity, potentially by February, you might be able to make a bot that does a lot. But does, as something in the way that you do it. But that requires you to be able to put down your thought process, your considerations, things like that. So I don't know if companies are working on making that easier, asking you questions.

 

00:29:33:01 - 00:29:40:07

Laura Gemmell

Oh, I've just got a business idea off the back of this, but never mind. So by February I will be doing this, no I’m joking.

 

00:29:40:09 - 00:30:13:22

Joyann Boyce

A whole other business. But no, that's really interesting because the potential is huge I think for marketeers and special, and specialists is still the same thing. Getting people to put their thoughts down is the harder part. Getting people to describe what they want, getting people to really emote. So it's not the same, but it's similar. I saw a TikTok, TikTokkers, where dancers were talking to each other in dance speak and they would be like, give me a boom kak boom kakkakak.

 

00:30:13:24 - 00:30:37:09

Joyann Boyce

If you're not a dancer, you're just like, that's what language is that? What are you saying? But they just understood each other. And I feel like in a lot of industries we have our own shorthand that is so niche in particular, like for us, when we're training clients, we have an understanding of the client's level of understanding of marketing, a level of understanding of inclusive marketing.

 

00:30:37:13 - 00:30:57:05

Joyann Boyce

And then that's how we shape the content. But it's only when you speak to them and you hear them describe something in a certain way, you must have it with AI all the time. When someone comes to you and says, oh, robots are going to kill us, or they come and ask you a specific question about Transformers. It's a very different level.

 

00:30:57:07 - 00:31:20:08

Joyann Boyce

So. Huh, okay, so that's February. Now say there's no regulations because currently there's chat about there being regulations, in five years where, where do you think, being optimistic as well, do you think AI will be in five years?

 

00:31:20:10 - 00:31:48:07

Laura Gemmell

I think I'm actually a little bit like not a pessimist because I don't think AI is a bad thing, but I think we will hit a wall with how fast the acceleration is going because that is how things tend to have happened previously. So in the past things have been restricted by computing power. I think this will hit a wall where it won't be able to run fast enough anymore because we need to feed so many words into it to get it to do what we want.

 

00:31:48:07 - 00:32:11:24

Laura Gemmell

So if I'm thinking about five years in the future, to me, I can't really picture a world where things are that different. Like maybe the barrier to entry is a bit lower. That's the thing. I'm really obsessed with actually, for people being able to use them, I think a lot of companies will have really accepted using chatbots and stuff like that every day in the same way that we use Google sheets and stuff every day.

 

00:32:12:01 - 00:32:37:23

Laura Gemmell

There is, the other alternative for everything keeps accelerating really fast and everyone could have, again going to Black Mirror, a little mini version of themselves that they've downloaded themselves into that helps them with their admin tasks all the time. I guess to me that's probably like the casual version of AI acceleration is everybody gets something that can help speed up their day.

 

00:32:38:00 - 00:32:54:01

Laura Gemmell

Out of it. And then I'm going to be really, really pessimistic and be like, but everybody wouldn’t get that. And actually what we're going to be doing is deepening inequalities throughout the society. But I don't believe this is the podcast for me to air those grievances.

 

00:32:54:03 - 00:33:13:07

Joyann Boyce

What I will say in a marketing sense, if we're not mindful, my pessimistic side is that marketers will be the fault for AI getting worse if it continues to scrape the content that we lazily publish without checking for bias and stereotypes. But any who on a positive note, we can all get a mini Miley Cyrus, that’s the episode you’re talking about? Yeah.

 

00:33:13:07 - 00:33:36:16

Laura Gemmell

No, actually that's a good but it's the same technology. I'm a real Black Mirror nerd. It's the same technology. There is a Christmas episode, I think it's called Black Christmas from like very early on back when it was on Channel four. And it's all about this green technology. And basically you have like a mini little clone of you that feels feelings and gets upset.

 

00:33:36:16 - 00:33:50:15

Laura Gemmell

But the whole point is it knows how you like your toast and stuff like that. So when it makes you toast, it makes you the perfect toast and like, it knows what temperature the shower needs to be like. Really random stuff. I do recommend watching that one. That's very interesting. The Black Christmas one.

 

00:33:50:17 - 00:33:55:21

Joyann Boyce

If there's anything to take away from this episode, watch all of the Black Mirror episodes and.

 

00:33:55:21 - 00:34:09:00

Laura Gemmell

As warnings, as warnings, not as instructions on what to do. I quite often see news articles and I'm like, you weren't supposed to take that as a let’s build this thing. You were supposed to take it as a bad, don't do it.

 

00:34:09:02 - 00:34:36:01

Joyann Boyce

But I think it was the one, I don't know which came first. The Black Mirror episode of the Boston Dynamics Dog. And they made a dog similar to one of the episodes and everyone is just like no. No. And then a couple of years later, they made a pledge that none of their robotics would ever be armed. And it was like, I hear you, but someone could just take your technology and do that.

 

00:34:36:01 - 00:35:01:03

Joyann Boyce

So you're still making the thing. But any who on a positive note, I think it would be rally good for marketers to have mini versions doing little things. And I like your tip about if you're going to publish it on the internet anyways, don't worry too much about it. Like that's I feel like that's a good bit for where we are right now in the whole IP situation of AI.

 

00:35:01:05 - 00:35:02:22

Joyann Boyce

So many said that she.

 

00:35:02:23 - 00:35:21:20

Laura Gemmell

Going to say another thing that I think I totally forgot to mention, that it might be very helpful for marketeers, is how easy it's becoming to transform content from one type to another. So the minute you make blog and yes you can use some software to make a robotic sounding voice, turn it into a podcast or a video that people don't want.

 

00:35:21:22 - 00:35:38:23

Laura Gemmell

But we are getting closer and closer to being able to like very quickly make, it could be deep faked, but if you're deep faking yourself, I'm not sure that's a problem. That's where my line is. If you're faking yourself or you have the literal written consent to deep fake someone in a really specific way, that's also fine.

 

00:35:39:04 - 00:35:54:09

Laura Gemmell

But don't do it in any other way. And like, yeah, like if I could deep fake videos of myself to make more educational content for our platform currently, I'd be really into that, be really creepy, but I'd be really into it. And I think that helps marketing as well, because then it means you can focus on the thing that you're really good at.

 

00:35:54:10 - 00:36:28:21

Laura Gemmell

So maybe you're a really good writer, but you're not amazing at video production. And yes, you should still always hire people who are. But sometimes when you're a one person band or you're the only marketing person for a company, you need a little bit of a helping hand. So that's kind of the opposite of my concerns about it. It's more of a levelling off the playing field in that respect because it's giving tools to people who might not be able to do everything themselves, which I think is really helpful for marketing, particularly because content is currently like so important for companies and such a big asset to marketeers and it takes up so much time to make.

 

00:36:28:23 - 00:36:50:10

Joyann Boyce

Content is becoming such a monopoly and the amount needed as well has changed. Marketing five, ten years ago is very different to marketing today and amount of content you have to produce and the types and the formats and so forth. This podcast is even an example of that in a different shape of content. I didn't think I was going to be doing when I started the company in 2017.

 

00:36:50:12 - 00:37:11:13

Joyann Boyce

So it's a variety. Okay. So the podcast is called Marketing Made Inclusive. Can you think of a campaign you've seen recently, any type of campaign, I won’t specify to be like one that was diverse and not diverse. Let's see if we can make it inclusive.

 

00:37:11:15 - 00:37:37:21

Laura Gemmell

I'm going to give you an example that is probably really annoying. So I'm quite deeply into building a tech platform at the minute. So most of my life is spent on the internet researching different software that can help me. And so there was a campaign by a company called I think it's pronounced Supabase, it's like a S U P A base, and it's a database on an authorisation management company.

 

00:37:37:23 - 00:37:59:12

Laura Gemmell

And their big tagline was building a weekend scale to millions, which I was like, oh, that's, that's very interesting. And I started using them now because they also offer, like their other thing is I don't really understand this. They're an open source version of Firebase, which I believe is like a Google product. So I guess they're marketing one.

 

00:37:59:15 - 00:38:19:01

Laura Gemmell

They're definitely marketing to me, a company founder who wants to do things fast and they're also marketing unknown brand name as a comparison, which I thought was quite interesting. I'm not sure if, where they are in erm of inclusivity, I’m just aware that they’re not marketing to non-tech people at all.

 

00:38:19:03 - 00:38:37:21

Joyann Boyce

Yeah, they're being very specific in their language, but it's interesting the weekend to billions line is still a very specific individual to word it because it's someone who has a company, one, is a CEO doing it themselves but also has a weekend.

 

00:38:37:23 - 00:38:55:19

Laura Gemmell

Yeah, actually that's a good point. Someone, yeah, actually you're right. It's marketing to the very typical like, oh, I've got an idea. I'm going to have a minimum viable product in a couple of days. Oh, I can do this at the weekend. I don't have caring responsibilities. I don't have a part time job that I have to do on top of my other job.

 

00:38:55:21 - 00:38:57:24

Laura Gemmell

Yeah, no, you're right. I didn't really think of that.

 

00:38:57:24 - 00:39:17:12

Joyann Boyce

But in that case, I think it's good because then it's removing the people who have more time because I guess they don't convert as fast as well. So that's interesting it’s, I wouldn't say it's inclusive, but I think it's intentionally exclusive because, was there any imagery with it? Was it just.

 

00:39:17:14 - 00:39:36:12

Laura Gemmell

It was like a black screen with like slightly light up dots in the background and then like green text. Yeah, it obviously really stuck in my head. It really converted me. I'm quite bad at like actually agreeing to use software for the platform. It feels like a really big commitment to me. And so this is one of the only ones that I've been converted to.

 

00:39:36:16 - 00:39:47:21

Joyann Boyce

No, I like that because then, because the only thing I thought I can go off with is if they had any kind of representation of the typical coder or the typical founder.

 

00:39:47:23 - 00:40:09:08

Laura Gemmell

It's very clean. Yes, I can say the whole website's very clean. It's just text. It's just yeah, like images, yeah. I would say that the brand isn't super inclusive because it's they're like, what's the term like? Blogs aren't super easy to follow along with their get started stuff. I'm like, this isn’t very beginner friendly but that's not that's neither here nor there.

 

00:40:09:08 - 00:40:14:00

Laura Gemmell

I think they're advertising, like you said, to a particular type of person.

 

00:40:14:02 - 00:40:27:21

Joyann Boyce

And to be honest, that particular, when I think about that persona, they may go to videos anyways so they can fast forward to the bits they want help. So it'd be interesting to see how many types of content they have for that persona they're targeting.

 

00:40:27:23 - 00:40:53:15

Laura Gemmell

That is interesting. So I don't like videos, so I never think to look for videos, ever. I was straight into it and I guess actually for following along with the tech blog, actually it's not as likely to go to a video. If you want to do something specific, it's better to have code snippets that you can copy and paste into something. But just like things like some things didn't quite work and if you didn't know how to debug it, you would probably be really demoralised and give up.

 

00:40:53:17 - 00:40:58:20

Laura Gemmell

Whereas I'm the opposite and I'm like, no, what's wrong with this?

 

00:40:58:22 - 00:41:19:02

Joyann Boyce

No, I like that. I like the cleanness of it and I think the campaign in of itself, the way it sounds, it sounds like they were intentional with that. Because they could have done a stock image, they could have done a typical, you know, male dude at a keyboard, coding away on the weekend. But they had the same persona but just made it universal.

 

00:41:19:06 - 00:41:45:20

Joyann Boyce

Someone who has a weekend to build and launch something that is what those people who are fast turners. I like that. Inclusive and exclusive. A good, good, ad. For those who don’t know, Laura and I was speaking before trying to figure out what campaigns have existed recently. And yeah, I'm curious to know what ads you get served, I’m curious to see your algorithm.

 

00:41:45:22 - 00:42:01:13

Laura Gemmell

Yeah, I've got to say. So most of, when I think of advertising, I'm currently watching Bake Off, so they got adverts in between Bake Off and they're really annoying and they actively make me not want to shop. Because in one episode of Bake Off, you get shown the same adverts, three or four times and it really annoys me.

 

00:42:01:13 - 00:42:13:01

Laura Gemmell

So like my current ones are Marks and Spencer’s advert, Domino's, and then like something to do with a car which like I’m not really into cars so like either, you know, and they're just so annoying.

 

00:42:13:03 - 00:42:15:17

Joyann Boyce

You don't like the Marks and Spencer ads.

 

00:42:15:19 - 00:42:19:18

Laura Gemmell

Oh, I liked it before I got me to watch it four times in one.

 

00:42:19:20 - 00:42:26:19

Joyann Boyce

But they're so, they're so cute that they're like our favourite. We do a recap of all the Christmas ads end of year.

 

00:42:26:21 - 00:42:57:00

Laura Gemmell

Oh, you should add this one and find it for a pub in Enniskillen in Ireland. And it's really cute. It's a little man who's like goes and puts flowers on, what you assume is his wife's grave and he's walking down the street in Enniskillen and no one's talking to him even though he's trying to chat. And then he goes into the pub and gets his pint of Guinness, obviously because he's in Ireland and a little puppy runs over to him and he plays with the puppy and then the people that own the puppy and like ask if they can sit down.

 

00:42:57:00 - 00:43:12:09

Laura Gemmell

So then they end up chatting and having a nice time. And I think the tagline they put on it was where there's no strangers, only friends you haven't met yet at the end, which I think is really nice for people that are lonely at Christmas. Yeah, I really liked it.

 

00:43:12:10 - 00:43:37:15

Joyann Boyce

I really liked the. Over the recent years, I found Christmas ads have been adding generational levels of awareness of like the loneliness around the season. It's really lovely to see that side of it. And obviously that's still selling something. But the Marks and Spencer’s one going to check that one out. But before we go, Laura, please let the people know where they could find you on the Internet.

 

00:43:37:17 - 00:44:00:05

Laura Gemmell

Well, I'm quite active on LinkedIn, both on my personal account and Taught by Humans account. I will say I'm quite active, so like only follow me if you want to be messaged many times. I'm joking. And also Taught by Humans, so it’s Taught By Humans dot com. Oh and we have a YouTube channel with some data and AI content on it also under Taught By Humans if you're interested in checking that out.

 

00:44:00:07 - 00:44:21:24

Joyann Boyce

We’ll add all those links in the Shownotes. Thank you so much for joining me for the Marketing Made Inclusive podcast. I have a huge favour to ask you as a listener or watcher. Please subscribe and recommend it to one of your marketing friends. That's the only way, I have a mission to make inclusive marketing the industry standard by 2030 and I need your help to do that.

 

00:44:22:01 - 00:44:31:09

Joyann Boyce

So subscribe leave a review and share it with your network. See you next week. When we'll continue to discuss everything Marketing Made Inclusive.

Previous
Previous

Ep 41: End of Season Two Reflection

Next
Next

Ep 39: Scaling Inclusive Marketing to a Team of 600+ with Bianka Faustin