Where did it all start?

AI as a field of research, in today’s modern terms, has existed since the mid-1980s. It was only in the early 00s that the field began to gain momentum. During the 10th century, people started talking about machine learning and deep learning to teach computers to solve specific problems.

Other terms such as artificial neural networks have been used to describe the approach to teach computers to solve problems by themselves. This is done by providing basic information, which can be seen as the tools, and then conveying the expected outcome. After that, the computer runs X number of tests to learn how it can ultimately replicate the expected outcome with the tools it has been given.

Generative AI

The generative AI that we see today, mainly represented by ChatGPT, is based on the technologies that’s been developed within the machine learning field since the mid-2000s. The ultimate goal has been to teach computers to communicate in a language that we understand. It was only in 2021 with DALL-E that a major step was taken, and an incipient high-quality and practically applicable generative AI saw its light. With the launch of ChatGPT 3.0 and 4.0, further steps were taken towards a very useful generative AI. This type of AI application is referred to as Large Language Models or LLM for short.

What is Generative AI?

Generative AI are AI systems that are capable of generating text, images, video and other media in response to a question or command. The system improves over time through more questions, more answers and the user’s quality assessment of these, while it also learns from new data that is inserted and can be used as basic material.

A picture generated from the phrase “A happy golden retriever with a bone in its mouth” on freepik.com

 

How can generative AI be used in digital marketing?

We’ve only scratched the surface of generative AI’s involvement in digital marketing. The various tools available today can be used in everything from ad texts with image material to sound and video. Using generative AI to help write efficient scripts or excel formulas is also possible.

The possibility of sharing one’s own data and letting the generative AI use it as a basis for solving the questions we as users put to it can in the long run provide insights we might not have been able to draw ourselves. At the same time, it is limited in its reasoning based on the given rules it has to deal with. This is especially true for a tool like ChatGPT. Let’s look at how generative AI can be used in each individual field.

Generative AI and SEO

ChatGPT and content creation using a few simple prompts is the first thing that comes to mind when it comes to SEO and AI, but there are a plethora of tools that base their SEO recommendations on AI. You can get help with everything from simple keyword analyzes to content production, gap analyzes and some coding. As with all tools, you should carefully evaluate which analytics, content creation or coding you trust and want to implement. Many times you need to evaluate and modify the analyses/texts/code you get so that it matches what you want to communicate externally but also internally. The time has not come to fully rely on AI yet, but it should be seen as a much stronger tool than we had ever before and which frees up time for other, perhaps significantly more enjoyable tasks.

Generative AI and SEM use cases

The way we search for information is about to change fundamentally. Both Google and Bing have active beta tests underway for the next generation of search engines. We can therefore assume that a lot will change for search engine advertising in the near future.

Within SEM, there are several different uses, and Google already uses a form of AI to be able to deliver relevant ads to the right people at the right time, primarily through its Performance Max and Dynamic Search campaign types.

At Kvantic, we know that data-driven testing is critical to success over time, so we experiment with generative AI to find new ways to optimize and improve our existing ad campaigns. For example, tests are carried out where we let generative AI find search terms based on product descriptions at scale. We also let generative AI assess the quality of ad texts we have produced ourselves and suggest improvements.

Generative AI and Paid Social

The potential for generative AI in paid advertising on social media is great. Meta today uses its AI model to deliver ads based on campaign goals to the right people, but has also added functionality that should increase the perceived quality of existing advertising material.

We advise our clients to test the AI ​​tools available for text, image and video to challenge their own creative ideas. Tools like ChatGPT, Pictory and GetIMG are worth trying, to name a few.

What limitations and obstacles are left?

More and more digital marketing agencies are profiling themselves with AI-based solutions. At Kvantic, we use available AI tools on a daily basis in our work for the development and optimization of digital campaigns. We see advantages in having a working methodology that does not solely put the outcome in the hands of AI. Where we are agile as development of AI within the marketing space moves forward. We have on numerous occasions faced and outperformed other agencies’ AI based tools in regular tests and, at this point,  continue to believe that our approach will generate greater value for our clients.

We are still in the early days of what generative AI can bring us as marketers and generative AI and AI as a field will develop rapidly in the coming years. Also, the term artificial intelligence is getting watered down by the day. It is common for artificial intelligence to be confused with other machine learning technologies, but also that the outcome of inquiries gets better as soon as you get more data to work with. That doesn’t mean it’s artificial intelligence. 

Final thoughts from us at Kvantic

It should also be remembered that ChatGPT but also other commercial tools are reactive. They require us to give them a set of rules to abide by in order for us to get something out of the information they generate for us. Sometimes it doesn’t understand what we mean and sometimes it simply can’t deliver at the level you might expect. A common example of this is that ChatGPT still cannot count how many characters a sentence consists of or is limited to a number of characters to respond to a prompt. We are already seeing attempts to make ChatGPT act reactively based on historical data and it is probably a matter of time before the personal assistants (Alexa, Google Assistant, Siri, etc.) start to be based on generative AI.

Then we have the algorithmic bias problem. Because the AI solutions are trained predominantly on anglo-american data from western countries the biases and cultural biases that exist within these countries get enhanced by the AI and applied to all who use them. The result being that cultural differences between countries do not get accounted for.