Hello, and welcome to Marketing Blabs. This podcast is brought to you by Marketing Labs, an expert digital marketing agency based in Nottinghamshire. If you're a business owner or a marketing professional looking for straightforward non-salesy tips and advice to help grow your business online, this podcast is for you. Strap in, because we're about to reveal the things that other agencies would rather you didn't know.
Hello, everyone. We are back, albeit after a small hiatus where the entire office has been riddled with flu. You'll have to forgive me because my voice is still a little bit tinny. I must sound like a Yorkshire version of Barry White, not necessarily a bad thing. For those that don't know who I am, I'm Tom, Creative Director here at Marketing Labs and your host for this podcast. In this episode, we're going to talk about some new and exciting digital trends, as well as how you can start utilising AI to help develop and grow your business. Today with me, I've got Matt, the CEO of Marketing Labs. How you doing, Matt? All right?
Very good, thank you.
And also with me I have Nick, our Head of Digital. How are you doing, Nick? All right?
Very good, mate. Thank you.
Good. Good to have you on your first pod.
Excited.
Yes, let's get going. And we've also got Josh, our web development guru. How are you doing, mate?
Pretty good, man. Happy to be back.
Good stuff.
Love the title, Joshie.
Guru.
Guru.
You are a bit of a guru though, aren't you?
I like to think so. I'm glad you guys do, too.
I suppose we better get started then, hadn't we? So what's new in the world of digital marketing then?
Analytics.
Oh yeah, GA4.
Universal Analytics has about a month left, two months left. So GA4 will be taking over.
That's a bit more in detail I guess, is it? GA4. What's the play?
It's a bit annoying to be honest. It's quite frustrating. It's a very different setup, technically quite different, fiddly. If you're trying to do it yourself it can be very fiddly, very complex. And it's basically a move to event-based marketing. So it fires based on events that you set up as opposed to universal analytics, which previously was just a kind of fire and forget setup if you like. It was very basic, but kind of thorough at the same time. And fiddly. Fiddly, GA4, so if you're not doing it now, definitely put some time and effort into setting it up before June because if you're running late and not quite getting there in the setup, you'll probably struggle. You'll probably find difficulties and teething issues that you'll need to go through and try and eradicate. And it can be very frustrating.
There's a learning curve as well, isn't there? All of the menus have changed. How certain things are described has changed. And even during setup, in universal analytics, the reports you're given are just off-the-shelf reports that anybody can readily find. They're available in every account, but with GA4, you basically have to build those reports out yourself, which is time-consuming. And also you need to know what reports to build. But it's great for blog posts because there's loads of websites that are talking about different types of reports you can build and how to build them, and it can be quite valuable for certain businesses. So it's about what kind of reports is right for you as a business.
Talk to me about calculus marketing because that's something that's come into play massively this year. How's that going to affect everything in the world of SEO and digital and tracking?
It's huge. Yes, big, big change. Probably one of the biggest changes in the internet for a long time. Traditionally, how somebody would be tracked online is you land on a website and that website would drop a cookie into your computer and it effectively tracks some of the things you do. Without cookies, you are in a world where generally you can only really track first-party data, so it's data on that website, as opposed to tracking between websites. So it gets much more difficult. But this is why it's so important to have ownership of who your audience is and know more about how people are behaving, and segmenting the kind of people you're targeting. It's going to be so much more important doing that, going forward.
I guess another big thing, we touched on it a little bit in the last pod, but TikTok is becoming more and more prominent now. TikTok marketing itself, that short-form video format is quite big at the moment. Do you see lots of people trying to implement that and not necessarily for the right reasons as well?
As ever with these trends, there's a huge amount of people who jump straight in. They don't really do it with a plan, but it's the new upcoming thing and it's exciting and it's working for some of them, and others, there's just maybe a bit of noise, but clearly TikTok's going places. It's doing things.
It's here to stay, isn't it?
Yes, for sure. It's here to stay.
It seems a lot of other platforms are heading that way as well. You look at YouTube and Instagram, they're all doing shorts and reels. It's definitely taking over, I think.
The influence it's actually had on other platforms is quite staggering. Even Instagram themselves admit that actually, if you really want performance on Instagram, it's all about reels. And that's taken straight from TikTok's playbook.
It's more engageable, I think, compared to posts. Especially thinking about Instagram particularly. I don't know, people tend to engage with that short-form video format.
I think it's another extension of everybody knows, who's been in this world for a while, everybody knows that attention spans are getting shorter and shorter and shorter, and it plays into that. You've got quick five 10, 20-second videos instead of five 10-minute videos.
Do you think things like that is going to, probably a question for you, Nick, more than, take over the more traditional Google Ads or YouTube Ads, that type thing?
I think eventually you'll certainly challenge that. And it's quite interesting actually, because my kids are getting to that age now where that's becoming kind of important to them. And the influence that TikTok has over them and how they search and how they look for things has certainly changed versus what it was five years ago or 10 years ago, or even 20 years ago when Google was kind of just breaking through and search engines were just becoming a big thing. So yes, without doubt they will look for things on TikTok now. It's slightly different. The intent's not quite there as it would be for search. But yes, I think it's going to be a big challenge to Google.
I saw something a few weeks ago where people are now using it... There's data to suggest that young people are using it specifically in America, where they might search for restaurants in a certain area and they're looking at reviews and it's short-form video, user-generated content reviews of certain restaurants or places to go, cafes, that kind of thing. So clearly, like Nick says, I agree. I don't think the intent maybe is quite there in terms of what people are searching for, but I think over time that probably will change and become more commercial.
Where is that being used? Like what platform are those reviews on?
TikTok.
So you're searching for your reviews actually on TikTok?
Yes. People are looking for things like coffee shops in X. It's mostly an American phenomenon at the moment, but I'm sure it'll catch on over here. And obviously TikTok is then presenting them with user-made videos about people in restaurants having food or-
I think certainly for lots of different types of searches or different intents that aren't quite specific to searching for a product. But lots of people use TikTok for that already. And I know definitely I mentioned earlier my two kind of do that. And one specifically recently, an example of that is where Harrison was looking for new LED lights and he was looking across TikTok to look for different people, what setups they've got or what products they're using, how they do it, how they instal it. And TikTok is the go-to place for that now, especially for younger people. And I think that's quite interesting because if they capture that audience now, in five or 10 years where they become buyers, that's going to be embedded. And that's definitely a big shift from where we were a few years ago.
It's that first-mover advantage, isn't it? If you think your audience is on TikTok, yes, absolutely. See if you can start capitalising on things like that.
It's interesting to see. And hearing you mentioned about, I guess I'm going to call it tech reviews with the LED lights. That's quite interesting because for a lot of my life, that type of media has been on YouTube and it's been in long form. And seeing that shift, personally, I don't know if I'm ready for it because I don't feel like you get the kind of information that you'd want in a short form, but I suppose the audience seems to be there for it.
Well, this what's interesting though, because actually what I've noticed myself, and I don't think I've seen any studies or any data to back this up, but it's just a feeling I have is actually there's a lot of videos on there that aren't even very accurate. It's just information for information's sake. So there's certainly an element of trust there that I think YouTube has that TikTok doesn't. As to whether that will change, who knows? It's very easy with the app just to have a quick video, a 10, 20-second video on TikTok.
There's a lot of kids doing it. Whereas YouTube's a bit more complex than that, and to get reach requires quite a bit of authority and expertise. I wonder how much that's going to play a part on whether TikTok's got anything up their sleeves that are going to improve search results and make it a little bit easier for people to trust the information.
Lots of things to think about there from a trends' perspective in digital marketing. But one hot topic at the moment out there with regards to trends is AI, Artificial Intelligence. That was a big word for me.
The holy grail. He founded that one.
A little, maybe.
[inaudible 00:10:11]. Which word, Tom? Artificial intelligence?
Both. But can anyone summarise what is AI and how does it work? Who wants to start?
I guess AI on its own is quite a general term because it can be boiled down to a lot of different specific things. You look at one of the small things that's happened recently is self-driving cars. That's a form of AI. It's a computer thinking about how to drive. But some of the more interesting things that are coming out at the minute are things like these language processing things. Chat GPT is a great example of that and that's using a neural network trained on vast amounts of data from all over the web to deliver you, I guess, useful information based on prompts that you give. That's quite a unique thing that Chat GPT does, but there's a lot of other things that are getting built on that.
I thought you were going to give me a massive spiel about Elon Musk then, when you said self-driving cars.
I could go into it.
No, let's not. You love Elon Musk as well, don't you?
He's a visionary.
What's your take on AI, Matt?
I agree with Josh. AI, it covers so many different things. I think what most people would consider AI at the moment, and maybe perception's changed a little bit, is Chat GPT. I think that's the one thing when people talk about AI, I think that's what people assume AI is, Chat GPT. There's a lot more to it than that, but obviously in the last three or four months, that's gone absolutely bonkers.
It's been mental the last few months.
Yes. Crazy.
Everyone is talking about it. LinkedIn.
Nonstop.
All over the place.
I've been playing around with AI for quite a long time. I know I mentioned in the first podcast, but I wrote a blog post, or say I wrote a blog post. I used AI to write a blog post about why you should use AI to write blog posts. And that was maybe a year or two ago.
Did that not mess it up?
No. It was a reasonable blog post. And this is the interesting thing. Even back then it was quite powerful. I think the one thing that's changed outside of just how powerful Chat GPT is, or the latest version of Chat GPT, is the natural language processing has clearly improved a huge amount, both input and output. So the things you input into it that it understands, but also the output that it provides, that's really moved onto a new generation of AI.
I think that's the most mental thing for me. I've only touched the surface of it, using it and the platform, but how the system itself is conversational and you can have a conversation with it and further optimise what you've asked it. And that's the bit that I just found crazy, really.
And it learns as it goes. It learns from that conversation. So the more you input as the conversation evolves, the more it learns from what you want from the output. So you can keep tuning and the inputs that you provide help it give you the answers or the output that you want.
To a degree, you're kind of programming it whilst you're speaking to it?
Yes.
In a lot of use cases, for example, there's been a few things I've done with it, like to automate tasks that I'm doing and I can tell it to do this, "I want to give you this information and you give me something back," and you can tune it to the point where you don't even have to prompt it with an opening sentence. You just feed it information and it feeds you what you want, which is quite powerful, I think.
I think that in itself will become a really useful skill over the next few years. If you expect to get good outputs from AI with poor inputs and lazy inputs, let's say, then the output just won't be great. Whereas if you really take the time to be precise about what you want from an output, the results that you get are significantly better. And I think that will become a skill.
There are quite a lot of tips going around at the minute, which I find really interesting, about developing prompts for the AI. One thing I've noticed specifically when using it to help with coding and writing scripts and stuff like that. It'll cut itself off every now and then. It'll spit out maybe half of a JavaScript file or something like that and then it just stops halfway through. I'm not too sure why that happens, but I know there are some tips and tricks that you can do to prevent that happening, so I've read recently, I've not actually dove into the article yet.
Interesting. I wonder if that would be the same on a paid account, uninterrupted? It might be a technical issue possibly.
Yes, it could potentially, like a heavy serve load or something like that. It would be worth looking at.
I think it's well worth paying for, personally.
How much is it, just out of interest?
$20 or something.
It's in a public preview at the minute, open public preview. So it is free to use, but that's not going to last forever.
Yes. If you want uninterrupted access, I think it's $20.
A year?
A month.
Oh, a month?
A month. I'm not quite sure exactly what that comes with, what benefits, but I know you get uninterrupted access when it's busy. And that's one of the biggest issues with it at the moment is very frequently if you're trying to use it on the free account, it'll just say it's too busy because it's blown up and it's gone viral and everyone's using it. But the paid account, there are no limits when it comes to that.
Talk to me about some specific uses, then. Nick, have you used it for what you do at all?
I have, yes. I'm going to be a little bit controversial, Tom.
Cool.
Because I think a lot of it at the minute is just noise.
Really?
Yes. And don't get me wrong, I think it will become something that's probably fundamental to what we do in a few years time. And I think it's starting to take those baby steps to get there now, but I think a lot of what it produces is still not that great, to be honest.
Quite basic.
Not that it's basic. I think you still need to do a lot of fact checking, so I think it depends on what you're asking it to do as well. I think there are certain use cases where I think it's really, really, really interesting, really useful, but you've got to put your name to it. Whatever the output is, you've got to put your name to it, so how much do you trust that that you're getting back and what does it look like?
And I think certainly when I've tested it, a lot of the output of it's still not really been that great. It's not really saved me much time on doing certain things depending on what the circumstances are. I'll give you two examples. One of them was good, one of them was bad. But in a recent example, I asked it to create headlines for some copy that I was looking for a client, just to see what it would do, what it would change, how it differs from what I do.
So I put in the instructions with the character limits, that sort of thing. And a lot of what came back was just nonsense really. So it wasn't very good, and then tried to refine it a little bit better and eventually got it to a state where it was okay, but it certainly didn't save me any time and it wasn't better than what I could have done.
And then the second example where actually I thought it was really good was when I was looking to build out some target audiences for a different client. So pulling together a view of affluent geo areas, so geo-targeting, and looking to understand what areas in the country actually hold a more affluent audience, if you like. So people who live there, either high property value or more senior job titles, whatever it might be. So I asked it to do that and look for top 20 different cities, postcode regions and that sort of thing to pull together a list of where the affluent areas in the UK are.
Interesting.
And actually it did that and it pulled it back and it came through really, really well. I did a little bit of research on it and it all seemed logical. So I think it depends on very much what you want from it, not necessarily how you approach it, but I think that does play a big factor as well. But I think naturally, at this stage of development now, it is better at doing some things than other things. Whereas in the future I'm sure it'll become fundamental to how we operate and how we work, not just within our industry, but lots of different industries as well.
That's really interesting. It's still really new though, isn't it?
Yes.
But I know what you're going to say because you're completely opposite to Nick I guess, in a way.
Yes, I am. But I see what Nick's saying as well though because I think there are certain, I guess, inputs that you would give it where it would struggle. And I think being openly creative is probably one of those. I think where it will improve, I do agree with Nick, outputs for things like that probably aren't going to be great. I think it's the kind of thing you've got to test, but it's almost one of those things that's like a lazy input. And I think when you expand beyond that into something that has a real use case like extracting data or using knowledge, I think that's where it really sort of moves into its own.
What I was going to say actually was I think its machine learning is now closed, it doesn't have access to the internet, it processed all of its data learning over a period of time. So it stopped learning as it is now. Obviously there's an element of learning during the conversation, but it's no longer learning from the internet. Now at some point, who knows when, but at some point obviously it'll open that up again.
I suspect the ultimate end goal for this is that it has access to information on the fly and learns on the fly. So, as and when that happens, so in a situation like Nick's talking about here, I see where there is potentially some huge improvements is where you could, for example, talking about ad copy and titles and maybe the descriptions in ads. There's probably a situation that you can easily imagine where you'd feed it a landing page and you'd say, "Now create an ad for this." And all of a sudden it understands the USPs, it understands the challenges, the problems that the product solves or the service solves.
And I think then the outputs should probably be better. But I also agree with Nick as well. Even though I'm an advocate, I also agree with Nick in the sense that 90% of everything you read about AI at the moment and all of the tips and hints that people provide, they're actually pretty generic. And I think most of them are quite lazy, to be honest. Most of them aren't groundbreaking. And maybe this leads on nicely because I was talking to Josh earlier about what Josh's ideal use case for AI was. And maybe this is a chance for Josh to talk about that because that sort of is groundbreaking and that is a use case that is actually really powerful.
Was that the content creation?
Yes.
I can't claim the idea as my own. It was actually something I saw on YouTube and it was a developer working with Chat GPT's API and they basically set up a Python script that spoke to the API and asked it to create some, I think it was inspirational quotes, based on their social media account. So they were a developer, they spoke about JavaScript and Python and coding and stuff like that. And it was coming up with quotes based on those kind of topics. And then it went as far to then create a prompt for another AI made by the same company, which is DALL-E. So Open AI owns both of them, but DALL-E is an image generation AI. And it created prompts based on the quotes to create an image of what that quote may look like.
Then downloaded the images, downloaded the quote as plain text, cropped the image to a Instagram story aspect ratio, like 16x9 or something like that, and overlaid the text over the image with a Python script, a different Python script, and finally published it to Instagram on the story, all automatically. And that for me was kind of incredible because you look at what people do at the minute with Instagram stories and I guess post to a degree scheduling stuff in advance, if you're a small business. This thing can do it on the fly. You can give it a prompt and it may take you 10 seconds to give it the prompt and then a few seconds later it's already posted to the internet. You don't have to do any research for the image yourself. You don't have to research the quote yourself. That was a quite powerful use case for it.
That's pretty mental that, to be fair.
It is, yes.
There's also a lot of trust that you're putting into that programme there. You need to really trust that the output's going to be spot on.
One thing I was thinking with that, and the developer I was watching was more or less doing it for the video content of the journey, as in creating this programme or the script that she was making and they had a fully automated process. What I could imagine and what would probably happen in a real-world scenario is that you'd put this together on a website or something, you feed it your prompt, and then it spits out a few different image options and a few different quote options and you can choose the best.
And you'd select which one you'd prefer as opposed to it just being-
Yes. And then even if you go a little bit further, like the option to regenerate the image but keep the quote or regenerate the quote but keep the image or something like that.
Yes, just add a layer of control in between for human interaction.
Yes, that also leads me onto something I wanted to say about what you were saying, Nick, and it was a little bit earlier on, but it's something that I touched on in the last episode as well. And I think when you talk about using AI and some stuff coming out as not that great and some stuff coming out as really good, I think it's quite dependent on a baseline knowledge of whatever the subject is in the first place.
I've been using it for coding quite a bit and having a good understanding of what you're working with already can be a lot more beneficial than just trying to do something from scratch. So I'm fairly confident with JavaScript and I can ask it some questions about JavaScript and kind of tailor the conversation in the way I want it to go. But if I'm working with a language that I'm not very good with, Python, for example, I don't really know Python, then I can ask it questions and it can give me some information, but I don't know whether that information is as good as it could be or whether there's a potential better way to do it.
Because there's been quite a few examples actually that I've seen online where people have inputted something expecting an output, and quite often the output is good and it fixes the problem. And we're talking specifically code here, but in some cases actually it's just downright wrong. So having that understanding of the code or the language used, you can eradicate that becoming a problem. If you were to just trust it blindly and implement it, it's not always going to work.
One thing that's quite interesting is in a lot of cases when you do trust it blindly and implement something you may get an error. Like say you're working on a website, you get a console error or something like that, you can then give Chat GPT the error and it can tell you how to fix that. It may take a few revisions, but it's quite good at fixing its own problems.
Has anyone got any? They were really good examples by the way, Josh.
Thank you.
I'm not saying that they weren't. I think coding's one of those things that's quite specialist. So have we got any examples for people who listen to this pod that could be quite useful?
Yes. So from a marketing perspective?
Yes. Digital marketing.
Yes. I ran a quick test this morning. I got an email from Costa this morning and the subject of the email was "Ultimate Blend is now in store." I look at that and I think, well, I don't know what Ultimate Blend is. I'm assuming it's a new coffee that they think is their best coffee. So I thought, I wonder what Chat GPT does with this? So I went to Chat GPT and asked it to recreate the subject line. That was a bit of a lazy input, the outputs I got were average. Slightly better actually than Costa's but average.
So then I inputted some of the information in the email and all of a sudden it opens up and you get some pretty nice subject lines. So instead of just "Ultimate Blend is now in store," I was getting outputs like "Immerse your taste buds with our new Ultimate Blend," and "Ready to try something amazing? Our Ultimate Blend is here." And they're subject lines that obviously are much more immersive. You get so much more engagement and the open rates on the email would be, I suspect, a huge amount higher.
So I think that's a really simple use case. Anybody that's doing any form of email marketing, before you publish that email, copy your subject, the subject that you think you're going to use, pop it into Chat GPT, give it some information about the email, the purpose. And I think the output will be much more powerful than the subject line that you're likely to pick yourself.
Did you go through a step-by-step process with it or did you just pretty much do, "Here's the subject line," what Costa gave you, improve it, basically, or did you get into more detail?
Yes. No, I think that detail's quite important actually. I usually start off explaining what the purpose is. So, this is going in an email. I need a subject line. Can you rewrite this subject line? This is the one that I've received. I think it can be improved. And you go on to, if you can explain how and why you think it can be improved and what the goal is, so "Increase open rates," your output is going to be much better. So it's all about the prompt. It's all about the prompt. It requires being quite conclusive.
It's only as good as the information that you provide.
Exactly.
One thing you mentioned there as well, which is quite interesting, I was playing about with this just the other day, is the importance of context. So giving it just the subject heading is okay. You can get a slight improvement from that, but as you gave it the body of the actual email, that makes a huge difference. I was playing about with it, it was social media post ideas and stuff like that. I was using it to make hashtags or a list of hashtags to go with a post. And on its own it did all right, relatively okay. But as soon as you feed it the actual topic of the post or the actual post body copy, it gets a lot more specific and granular, I guess.
I used it, interestingly, I wasn't going to mention this, but Josh has just reminded me about this. The blog post I wrote before last, I was struggling with a title for the blog post. I had a few titles, but I didn't really particularly think they were very good. And I fed that entire blog post, 2,200 words or whatever it was, into Chat GPT, and asked it to give me a title. And it was a good title, much better than mine. I tweaked it slightly still, but it's that inspiration where I think it can be quite powerful. And that takes 20 seconds, 30 seconds, and all of a sudden you're producing something that is actually better than you would've produced anyway.
This is my own opinion, I think the core purpose of Chat GPT and anything AI at the moment, is better for idea generation than it is the actual output itself, I think so anyway.
In certain cases. I think specifically with writing and creative stuff like that, specifically for me, I'm going to keep talking about coding because that's what I do. But when you look at stuff like that, I think the output of that is quite powerful. And the fact that most people wouldn't be able to do something like that. And it is to a degree, still not perfect. It doesn't come back with the greatest stuff every single time. But it's better at coding, I think, than writing text.
I actually used Chat GPT to write a song.
Are you about to get your guitar out?
I'm not getting my guitar out. I might get it out on another episode.
Episode 10.
I promise. No. But I can't remember whether I mentioned this on the last pod.
Yes, you mentioned it but you didn't go into detail.
I just basically gave it a brief outline of what I wanted it to be. Also, I said what key I wanted the song to be in, so key of G, for example, and this was my chord progression. This is what I want it to be about. Sadness.
So wait. Did it write the music or the lyrics?
No, the lyrics. Sorry. It was really quite amazing. Some of it was really bad. I would never say it in a song, but at the end of the day it's a song and for it to come up with the lyrics, a chorus, a bridge, verses, I found it amazing, to be fair.
Have you sang it? Have you tried it?
I can't sing at the minute, but I will one day. It went something like this.
I'm on the edge of my seat.
You had me then.
Everyone went quiet for a little second thinking, "Oh, Tom's going to sing a song." Just get me my guitar. No, I'm only joking. I will do at episode 10.
Deal?
Yes.
Commitment?
Yes.
I've got something I've been using it for as well on a personal level, non-marketing related. But when I put my little boy to sleep at night, for people listening, Leo's now three and he's at the age where he's starting to get quite creative and imaginative. So when we go to bed at night, instead of reading a book off his bookshelf, we'll specifically talk about what kind of book he wants to listen to. So I input some information into Chat GPT and I've got a long chat going. I've kept it in the same chat for context.
So Chat GPT has built up a bit of a character for Leo. He's a superhero and he's got a sidekick called Chester who's a dog. And obviously my dog's called Chester. And each story that we create every night brings the same characters back into play. So it's really nice. So Leo might say to me, for example, one night, "I want a story about tractors," and this particular tractor might be a rainbow tractor and it flies. And again the output from Chat GPT in terms of a story for a three-year old is actually brilliant. Leo loves it. He just beams the whole way through the story.
Does he give you the ideas for it, though?
Yes. We sit in bed and I'll say, "What do you want a story about tonight?" And he'll reel off just some ideas. It takes three or four minutes and a few little ideas here and there and I'll put them in and within seconds you've got a customised story.
That's actually a really cool use case for it. I wouldn't have ever thought about that myself. Obviously, I don't have kids.
Can you write me a story, Josh?
Yes, maybe. It wasn't going to [inaudible 00:32:08] well.
Dynamic duo who create amazing websites.
Does it sound familiar, that?
Yes, it sounds very familiar actually. I feel like I've lived it.
So Leo does love it. He thinks it's amazing. And all the way through he beams. And I think anyone who's got young kids, it's definitely worth trying because it's really good fun. They love it and they think you're sort of magic for being able to just create this amazing story on the fly.
Do you know what, as well? On top of that, it's kind of like an early induction to that now, isn't it?
Yes.
And kids of today are going to grow up doing this naturally and it's going to be part of their life. Three or four, five, 10, however many years, and by the time they get to the age where they're working and doing X, Y, Z, it's going to be-
It'll be the norm.
Embedded.
It'll be normal and fully embedded.
It's getting to the point now where it's like augmenting humans, I think, or it's definitely heading that way at least.
I think that's a bad thing, though. Don't you?
Maybe. It depends how it's used. If it's used to complement, supplement support, then it can only be a good thing.
I know. But then I just get this vision where everyone just will start relying on it. They won't be creative. I sound like my dad or something. "Well, back in my day, I didn't do that." But, I don't know.
No, it's true. We said earlier, didn't we, that attention spans are getting shorter and shorter? That's a direct result of the impact that social media's had on society over the last 15 years. Maybe there'll be another one like that for AI, who knows? But at the same point, if it's used correctly and appropriately, I think it should be a force for good. It should supplement what people do and support them.
The scary thing about it people can use it now to write their uni dissertations and stuff like that. That's the scary bit. And there will be people out there doing it, to be honest. There will.
There are lots of people doing that. And I think when you look at education, things are going to have to change quite drastically to get on top of this and make sure that I guess, to a degree, people are being graded correctly, or at least the way people are being graded is going to have to change to either allow this or disallow it.
But aren't you going to have to use AI to see if someone's article has been written by AI, eventually? That's how mental it gets.
It is. And this actually, I know we wanted to talk about responsibilities and potential issues with using AI-
Risks.
Risks. And this is maybe the first one. It is something to, I wouldn't say to necessarily be too concerned about, but it's something people need to adapt to. And just on that note as well, the biggest risk that I can see with specifically a marketing podcast, and I think one of the things I would urge for people listening is not to just take an output from Chat GPT and use that on your website. Fine, use it in socials. That's okay. That's not going to impact performance really. Or at least as it stands at the moment.
But I'm seeing a huge amount of data to suggest that if you're using it regularly on your website, it's going to be problematic. Google can trace it really easily. In fact, you can trace AI-written content from tools that you can find online that are free, that were trained on Version 2 of Open AI. And it's still picking up pretty accurately the outputs from the latest version of Chat GPT. So even though it's amazing for all of these things, you've also got to know what is safe to use it on.
I think, and to add to that as well, what's really interesting, Google updated their guidance, didn't they, around two weeks ago or maybe three weeks ago? And now for the first time they stipulate why and what has created the content. So it's asking you to inform people the purpose and how it was generated. That's very clearly one of the first indicators that Google are taking this seriously now in terms of rankings, moving forwards, how it detects and how it defines whether that content's useful I think is going to be a significant indicator as to the value that AI provides to website content, I suppose.
And interestingly enough, just for the people listening, Open AI is the AI system that Chat GPT is built on. Open AI also updated their terms and conditions recently and they're suggesting that actually all of the outputs are owned by the person that are putting the inputs in. That in itself is quite telling because it shows, I think what I take from that is that they can't be responsible for the outputs. So if there were any legal issues, any issues in terms of quality of output or in terms of consistency of output, things you can't say, things you shouldn't say, things that are going to land you in trouble, things that aren't accurate, they're basically saying it's down to you. It's your content.
That'd be right as well.
Exactly.
Who legally owns that and how much of it's copied from elsewhere.
And that's in their terms and conditions now. So that leads me to think that they're also conscious themselves that the outputs, it's not fact checked, it's machine learning and natural language processing trying to understand things. There can't be somebody there checking this output. So it's on the person inputting to check it. So if there are any facts in there, they need checking. If there's any statements in there, they need checking.
So there's a few things to be aware of. From an SEO perspective, obviously you've just got to be really careful with the outputs. Don't just take them and use them. As I was saying earlier, it's quite powerful if you use it to supplement your process. So if you want information on things, if you want it to help you create a brief for a piece of content or you want to align your audience with the content, great, they're really useful inputs and the information it gives you will be handy. But actually just getting it to create you content all the time, it's not going to be helpful. You'll probably be removed from search results pretty quickly.
I suppose the same kind of applies with what I do as well because if I was to take a piece of code blindly and add it to a site and that was to cause a problem on the site or maybe even be insecure or something like that, then it's not really great. And the fact that that falls back on whoever implemented that code, and not Chat GPT, is quite interesting.
Do you think Google's missed the boat a bit with Chat GPT? Because is it Microsoft who own Chat GPT?
They've-
Acquired it.
Acquired it.
It depends what you mean by missed the boat. Because Google have been doing lots with AI and machine learning for a long time. That's fully embedded into ads, for example, and how it finds audiences. So obviously this Chat GDP... I said that wrong, didn't I?
Chat GPT.
Anyway, that's obviously a different implementation of that. So how you use AI and what its purpose is, I think, has scope across lots of different products. So whether it's missed boat on implementing and integrating that into search, maybe. They've got their own version of that, I think. Probably not as useful at this point in time, but whether it will become more useful, probably a safe bet on Google for doing that.
And they still have the largest search volume data. So I think they're probably confident that if they come to the show a little bit later that they should still be fine. As Nick says, there's so many use cases for AI, it's not just Chat GPT. Google have been using AI in their algorithms for a long time. RankBrain, the algorithm, that uses a lot of machine learning and understanding natural language processing. That's evolved and more recently evolved into BERT, and BERT is incredibly powerful.
And that's all about understanding language and that's machine learning and partially AI driven, I believe. So it's more a case that all of that is hidden behind the scenes. When you're searching in Google, you don't know that's going on. You don't know that exists. It just happens. I think the difference here is, with Chat GPT, it's a conscious decision, almost conversational with AI. So the fact that Microsoft have invested or acquired Open AI and integrating Chat GPT into Bing, I think it definitely gives them first-mover advantage and I suspect they're probably picking up a little bit of market share right now.
It's whether that market share is sustainable or whether they're going to go back to Google. And also when Google properly get their AI usage into search results, what kind of impact that might have on things? I think if the demo that Google themselves use was anything to go by, it looks pretty poor because the information that they took screenshots of was inaccurate. So their big launch party, they showed straight away some of the limitations and risks of using AI. So clearly it's not ready for it.
Not ready for it, yes.
You mentioned that it's been used in ads and behind the scenes so much and we've talked about obviously it being used for text content generation and image generation, to a degree, but there are quite a lot of other use cases for it on there.
Yes. It's almost endless actually, isn't it? Even a few years ago, there's been apps that people have on their phone for a few years where you'll take a photo and it'll turn your photo into say a cartoon in certain situations. They'll put you in front of the pyramids in Egypt, and that's all AI-driven. They take the lighting, they take the shape, the shades, and they input it into a different background photo.
That's really cool. There was one that I found the user interface and everything was really cool. It's called SOUNDRAW, I don't know whether you've heard of it.
No.
Basically, it's an automated music generator that allows you to create royalty-free AI music for things like videos and content. You basically put in some basic parameters and decisions about maybe the genre of music that you want to create, the instruments that you want to use and the sound, the mood and the length of the track from say 30 seconds to three minutes, and it outputs a royalty-free song-
That's cool.
... That you can use. It is a paid platform, so you are going to have to invest in it if you want that. I think it was something like 20 quid a month or something like that.
All right. Do they have free trials?
They let you hear the song-
But just can't download it.
But I don't think you'd have an actual trial.
I'm just wondering if we could do one by the end of the podcast and show it.
Why don't we save it for podcast 10 and it can do the music for your song?
Amazing. Let's do it.
Let's do it. So there's a promise, episode 10.
You're going to sing over AI-generated music.
Yes. It better be good.
With AI-generated lyrics.
It better be good because I am a class at singing.
Debatable.
Not really. Another one quickly, example, just another quick example is a... And I don't really like speaking about this because it's design based, but there's an example called Looka.
How are we spelling that?
L-O-O-K-A. Looka. It's a tool, like I said, I'm reluctant to promote this, but through gritted teeth, it makes it easy for startup businesses to brand their business. It uses AI to create unique logos that convey the company, it gives you a colour scheme, fonts, everything like that and it gives you an overall visual identity. But it's really basic. It doesn't take into account your audience.
But for startups that kind of thing's quite useful.
Oh yes, for startups it's ideal. You wouldn't necessarily use something like that for a full brand refresh or something like that where you've got to take into account maybe a new market or a new audience or something. But through gritted teeth, it is a pretty cool little platform, to be fair.
Nice.
That's bordering on towards the creative uses of AI, which I don't think are quite there yet. You look at some of the image generation, some of it is quite good, but then a lot of it is very flawed, to a degree.
I did one recently where I asked it to create, I can't remember the specific prompt, but it was something along the lines of "Can you create X in the style of Monet?" And I ended up with a mouth where an eye should have been. So, like you say, it's clearly not quite there, but still very impressive.
Nice one. That's it from us then. Some interesting stuff that we've talked about there with AI, Chat GPT being the main topic, but there's been some good software examples and we'll put those into the pod description for you so you can research them and have a look yourself. But for now, thank you, you guys, for being on the pod.
Thank you.
Pleasure.
Cheers.
Take care. Look forward to my bedtime story, Josh. So thank you everybody for listening. It's been an enjoyable podcast. We've talked about how you can use AI, specifically Chat GPT, to improve your workflow for things like social media, content marketing, even writing music, or even a short story. Remember though, you are liable for the content that you output. It's still yours and you're only as good as the prompts that you provide. We've talked about examples such as DALL-E or Craiyon, which are great image-generation AI tools. We've also talked about SOUNDRAW, which is a royalty-free music platform, and Looka.
In our next episode we'll be discussing what it takes to create a great brand. Our team will share insights on key elements like brand messaging, understanding your audience, building trust, and creating a visual identity. Tune in to learn how to build a memorable and successful brand. Until next time, I'm Tom and this is Marketing Blabs. Thanks for listening and have fun with AI.