As an AI and Blockchain Expert, I was invited to an interview to discuss AI and digital marketing.
We discussed artificial intelligence applications in marketing and digital marketing in the next few years. We talked about recent trends such as social media influencers and issues related to that topic, fake influencers, bought followers, young generation following the wrong advice. We also touched on issues with global-scale digital marketing campaigns, such as negative comments.
We hope you enjoy this interview with Stella from Suceedigital.
https://www.youtube.com/watch?v=CVzbUKBP8iA&t=193s?embedable=true
Stella: Could you please tell me a little about yourself and what you do?
Me: I've been involved in data science for more than ten years. I did my Ph.D. in machine learning at UC; I'm still associated with UCL as a member of the Blockchain Center. Besides this, I'm also involved in various ventures. For example, I'm the CEO of the Tesseract Academy, an executive education company educating decision-makers on topics like data science and blockchain. I'm also a partner in Electa Consulting and the most involved in their startups' advisory roles.
Stella: Thank you so much for joining us. Today we're going to talk a little bit about how artificial intelligence can be implemented in the digital market. I would like to bring up the recent trends that we've got in the digital marketing industry; the first is influencer marketing.
For example, we are heading toward social media celebrities, but unfortunately, we have issues. We have fake followers nowadays that are being bought or created by both machines, and not everyone is aware of it, and not all companies are aware of it.
When they do research on a specific influencer that they need for their marketing purposes or that they apply for digital marketing campaigns, for example, they ignore their engagement on their analytics on how they're socially those influencer social media is performing, in general, that way they are spending they're devoting their budgets on someone who sometimes is just not going to bring a positive return on investment.
So the issue is that the young generation is mostly following those social media influencers, and they're following their advice which sometimes gets wrong due to social media influencers just wanting to promote the product they haven't tried. This means they give low-quality content, so I would like our viewers to become more aware of the issues we face daily as marketers and business owners.
How can you address those problems, and how can artificial intelligence, for example, help us to support those fake influencers?
Me: I think that's a great question, and here's a straightforward answer. Machine learning and AI have been used for a long time for fraud detection. For example, they can be used in the financial industry in banks, and there's no reason why similar techniques and methods can also be applied to influencers so we can support fake influencers to identify anomalies in the fan base.
It's straightforward to use an algorithm to detect whether someone's fan base growth is normal. That is, it closes very rapidly. For example, it's just the case that maybe many people in the world of marketing may not be as aware of those issues, or they might not be very aware of those techniques these techniques that are available to them.
Social media marketers come across, for example finding social media influencers. Then the companies do not pay attention to influencers’ analytics and engagement and some social media influence. I can have thousands of followers or even millions of followers but not have enough likes on my posts which is sad as companies are devoting a big budget to this.
Stella: I'm glad to hear that artificial intelligence can solve those issues; it's just a matter of us, probably marketers understanding how that can be used and how those processes or platforms can be used.
Me: In such cases, I believe that data science can help on multiple fronts, so again, talking about discovering fake accounts, and this obviously can relate to, you know to those accounts which you mentioned where there might be a large number of followers, but the posts are not getting enough likes.
Another question related to this is how you can measure engagement, so we have different social media platforms. Someone might ask the question: okay, so a like on Twitter is worth as much as a like on Instagram, for example, yep, and so statistical modeling can help us answer this question. We can also compare different influencers against each other and the norm. This can give us a better picture if giving all those tools together.
This can give us a better picture of where every influencer stands compared to the rest.
Plus, another set of tools I am very fond of is anything relating to forecasting, so predicting the growth of someone's fan base, the development of a post-engagement post we'd all get.
These are some exciting applications of data science in the context of social media and social engagement. Unfortunately, not everyone, and we're at this stage. Still, hopefully, through this video, we can send a message to people, to marketers, to business owners about how they can transform their strategy into something practical well.
Stella: Do you think we must start preparing to collaborate with IT as marketers?
Me: We never looked at that aspect in our industry because we have social media channels like Instagram, Twitter, and Facebook. Then we have programs like Salesforce and HubSpot that we use nowadays.
But do you think that marketers will start collaborating with the IT sector? Most professionals, and that's true for marketing as well, need to have some basic understanding or intuition about how algorithms work and how they can help them in their jobs. At least they would need to understand how well they can apply an algorithm so diverse that we can measure the engagement and compare influencers to generate content.
But also, they need to be aware of the limit of those methods because right now, you see in different industries either some resistance to change with a belief marketing was guilty of that for the last few years. But things are changing, or you see the opposite, like magic. So we think that the algorithms could do everything, so you can only get to the truth through basically practice education and basically through more education of the practitioners of a field. So that's a good question.
Yes, I believe that all professionals, market marketers included, will have to get more accustomed to working with machine learning. This is something to look forward to in the following years. It's a strategy that is applied in digital marketing. And yes, there were campaigns all flying back in time. Nowadays, running a campaign online is much easier because you reach the audience globally on an international level. If you are located in one country, you get such brand awareness faster in minutes.
However, for example, if we are running a campaign on that global scale, there can be some issues with negative comments that people write about the brand, and monitoring that in real-time is complicated because, for example, if that's just a simple market where have an eye job. But what? How do they monitor that outside of work hours? So can you tell us if there is an opportunity for marketers to solve the issue of negative comments we use? How do we avoid that issue nowadays?
So I think this can be solved through machine learning; it's a simple classification problem to detect negative or inappropriate comments. The challenge is the nature of this. They make that the exact nature of the words. What I mean by this is that they're, you know, it sounds tough to find datasets of, you know, with positive reviews and negative reviews and then create an algorithm that detects negative sentiment. But there are certain cases where creating an algorithm to see, you know, posts that shouldn't be there it's pretty challenging, and you still need humans.
I know for a fact, for example, that Facebook is still blowing a vast number of employees at screening content, maybe relating to terrorism, you learn to abuse wh,atever. They're also working on creating machine learning algorithms that can do this automatically.
So the challenge with a new campaign is whether comments of a very negative nature, whether there would be something detected by a machine learning algorithm, or whether the heart of the comments would be sad that the algorithm would be full to do everyday things.
So to give you an example, many algorithms in natural language processing most algorithms they have an issue with sarcasm. If any negative comments were very sarcastic, they would probably go through an automated filter without being detected. So there is a solution, but the devil is in the details again. It's more of a technical matter as to when you could apply an algorithm.
Stella: And how and under what circumstances well would it work? What's the best way to construct the pipeline?
Me: Starting from the algorithm can predict 100% a human natural language processing, right? The machine cannot predict that well; you never get a hundred percent accuracy. So you know that there are going to be some mistakes that are going to take place. Now, these mistakes will be 0.1% of the cases that an algorithm sees; will they be more like five or six percent, depending on the volume and the commons? So, in this case, you can make a massive difference in the final numbers.
Stella: Thank you for this helpful information.
This is the end of the interview. We hope it will be informative and insightful for any reader interested in AI, machine learning, and Digital Marketing topic.