Skip to content

In April 2023 I wrote an article called AI in Digital Marketing. Well, I wrote the second half of it. For the first half, I used ChatGPT version 3.5 with the prompt “Write an article that introduces AI and explains the benefits and disadvantages of using AI for digital marketing as if you were talking to a friend”. You can read that article here.

We are now 10 months on. AI continues to evolve. OpenAI’s large language model, ChatGPT, has a new version: 4.0! To test its evolution, I was curious to see how it would react to the same prompt that I gave to version 3.5.

But first, what are the improvements? What should we expect from the latest response to my query?

GPT-4 represents significant advancements over GPT-3.5 in terms of size, capabilities, and multimodal abilities. It offers improved linguistics, better information processing, improved creativity, more logical content generation, and more effective problem-solving skills.

Well, I guess a lot of that is lost for my simple one-line query..! The upgrade offers a lot of features that I’m not exploring today. My goal is to test a relatively simple query without any subsequent modifications, and then compare any core differences between the two outputs; specifically, differences that could be due to the improvements between version 3.5 and 4.0.

By the way, ChatGPT 4.0’s response to my query is impressive and worth reading. It is informative and correct. It is easy to understand and provides useful information. I recommend that you read it – it’s at the end of this article.

Remember, the point of my test was to test the default output; to test the starting point that is offered. It’s really interesting to see such different approaches for answering the same query.

So, how do the two different outputs compare?

  1. Both outputs were casual and friendly – but in different ways. The newer output uses metaphors, whereas the older one is more straightforward and slightly more formal.
  2. The source/training data has been updated. This is shown quite nicely by the newer version citing concerns of job displacement and talking about how cost-effective AI can be. Whereas the output from the older version states that AI is expensive.
  3. The newer version seems to offer a more balanced assessment between the advantages and disadvantages.
  4. The newer output is also, arguably, laid out in a clearer format (without any prompting to do so). It has been split out into easy-to-understand sections, with each having its own subsections. The older version only used paragraphs.
  5. Nevertheless, there are many similarities between the outputs too. Much of the core topics are the same, despite being worded differently: They both talk about data analysis and personalisation (Whilst also acknowledging it can also be impersonal), how it helps with automation, and they both talk about the dangers of over-reliance.

And here we go, this is the output from OpenAI’s ChatGPT v4.0 for that same query:

“Hey there! So, you’ve probably been hearing a lot about AI, right? It’s like everywhere these days, especially in digital marketing. Let’s break it down over a cup of coffee, just friend to friend.

What is AI?

AI, or artificial intelligence, is kind of like having a really smart robot buddy. It’s a technology that lets computers do things that usually need human smarts, like understanding speech, making decisions, and recognizing patterns. You know how Siri or Alexa can chat with you? That’s AI at work!

The Cool Stuff About AI in Digital Marketing

Personalization: AI is like a mind reader for your shopping habits. It looks at what you like, what you buy, and even when you shop, and then it shows you ads or product recommendations that you’re more likely to be into. It’s like having a personal shopper inside your computer.

Efficiency and Automation: AI is a time-saver. It can automate a bunch of routine tasks, like sending out emails or posting on social media. This means marketers have more time to think of awesome new ideas instead of getting bogged down with the boring stuff.

Better Insights and Analytics: AI is like a detective. It digs through heaps of data to find patterns and insights that a human might miss. This helps businesses understand their customers better and make smarter decisions.

Cost-Effective: In the long run, AI can save money. It’s like having an employee that works 24/7 without needing a break or a paycheck!

The Not-So-Cool Stuff About AI in Digital Marketing

Loss of Personal Touch: Sometimes AI can be a bit too robotic. It lacks the human touch, you know? People can tell when they’re talking to a machine, and it can make things feel less personal.

Privacy Concerns: AI is a bit nosy. It collects a lot of data about people, and that can make some folks uncomfortable. It’s important to balance personalization with privacy.

Dependence on Technology: Relying too much on AI can be risky. What if it breaks down or makes a mistake? It’s important not to put all our eggs in the AI basket.

Job Displacement: This is a big one. As AI gets better at doing human jobs, there’s a worry that it might replace some jobs. It’s something we need to think about and manage carefully.

Wrapping Up

So, that’s the lowdown on AI in digital marketing. It’s got some amazing perks, like making things more efficient and personalized, but it’s not without its challenges. The key is finding the right balance, using AI to make things better without losing the human touch. Like any good tool, it’s all about how we use it.

Hope that gives you a clearer picture! Let’s grab another coffee soon and chat more about this techy stuff, okay?”

And there you have it. (Is GPT-4 is obsessed with coffee?)

So which elements of the update from 3.5 to 4.0 have caused the differences?

It is clear to me that there have been improvements in linguistics, as evidenced by the metaphors used. I can also see more logic in the content that’s been generated, as shown by the arrangement of the output and the more balanced assessment. It’s great to see the improvements in action. I find that the improvements have helped make the content easier to digest.

Personally, I’ve found both versions 3.5 and 4.0 useful. In particular, I really value the multi-modal input capability of GTP-4 too, which wasn’t tested in this article. The multi-modal input has allowed me to input datasets that can then be tidied, analysed and represented, after which I can share my findings. I guess that’s a great example of “using AI to make things better without losing the human” touch. I perhaps do not use ChatGPT as often as I could. I could use it more, and perhaps that’s something I should look at. In a way, I’m practising the opposite of the overreliance that both the outputs highlighted!