Research 1:
Generative AI & Synthetic Media

Introduction

For my graduation project I will investigate how generative AI can be used within GroupM. The reason for this being that AI has the potential to automate processes within GroupM so they become easier and more efficient. It has lots of business potential and is for this reason important to look into possible future applications.

But before diving deeper into the subject I found it important to get a better understanding of what generative AI / Synthetic Media is, as I am new to the subject. By knowing the new technology I will use for my assignment will result in a better outcome of possible products. In this research I will be looking at the history of generative AI, existing tools and how they work.

Research strategy

Main question What is generative AI?
Subquestions
  • What is the history of generative AI?
  • How does generative AI work?
  • Which different generative AI are out there?
  • What can I create with generative AI?
  • Method used Literature study
    Date 09-02-2023
    Link to document Research 1: Generative AI & Synthetic Media

    Results

    Generative AI is a term to describe an AI that mainly focuses on creating new content based on existing data. This new content that is created by such an AI is called synthetic media. The GAN model was developed by Goodfellow in 2014. This new technological development inspired researchers to search for new possibilities with these generative AI models. This leaves us now that generative AI can create highly realistic images, helps with answering questions, create videos, audio, etc.

    The creation of synthetic media is most often done with a text prompt. A text prompt is a sentence describing to the AI what you want to have as an output. The prompt itself has a great influence on what the final results will look like. As mentioned before there are a variety of possibilities to create content with generative AI. The most sophisticated AI is now text-to-text and text-to-image. (Wurmser, 2023) However, as time passes other AI tools will become better. But for now, the most applicable tools right now are the ones that create text or images.

    Some problems and limitations however were found in the use of such technologies. Text generation can give incorrect and false information, is biased, uses vast amounts of computer power, has no insight into dataset and more. Except for the incorrect information, image generation also shows these problems. (McKinsey, 2023)

    Conclusion

    To conclude this research defines generative AI as a tool that can create new content based on existing data. The newly created content is called synthetic media. Meaning that everything created with such a tool is synthetic media. At this moment text and image generation is most applicable for GroupM. As these tools can already create realistic and usable content. However, it is important not to exclude other content types as these will become better over the years. To create such content it is done via a text prompt, a phrase that describes to the AI what output it should generate. The important finding being that the text prompt has great influence on the outcome. Further research could help indicate how to write a good text prompt. There are many possibilities with generative AI and some problems that arise when using it. For my next research I will look at the impact and possibilities this technology has on the marketing industry.

    Sources

  • What is generative AI? (2023, 19 januari). McKinsey & Company. Geraadpleegd op 9 februari 2023, van https://www.mckinsey.com/featured-insights/mckinsey-explainers/what-is-generative-ai
  • Wurmser, Y. (2023, februari). ChatGPT and Generative AI. Insider Intelligence, van https://www.insiderintelligence.com/content/chatgpt-generative-ai
  • Next: Research 2: Trend Analysis