Kasper Lindskow

From Generative AI to Prompt Engineering – Understanding the Concepts

Kasper Lindskow is the head of research and innovation at Ekstra Bladet and a postdoctoral researcher in strategic and ethical implications of AI in media at CBS. He is also a member of the board of the National Advisory Board [www.aicentre.dk].

Kasper will report and answer questions about the significance of AI for people’s work lives. Although it is easy to start experimenting with generative AI, mastering it is difficult. To improve your work with generative AI, you can read some of the many “prompt engineering guides” already available on the internet. They teach you how to work with the input to the systems to get the output you are looking for.

At the same time, it’s important to understand the limitations of the systems before you start using them professionally. Many systems have a liberal relationship with facts, know nothing about the world after the point they were trained, and completely break down when trying to work with quantities and numbers. If you are not aware of these limitations, your work with generative AI could end up worse than an unedited Google Translate translation – or resemble news from a Russian state media. It’s essential to be aware of your organization’s policy on generative AI, which should describe when and how generative AI may and may not be used by you. If your organization doesn’t have one yet, it’s time to create one, if only to be prepared for the developments that will occur in the coming years.

Right now, generative AI should be seen as a tool that can be used by a communications worker to write first drafts, generate ideas, correct spelling errors, etc. At the same time, you need to read everything that comes out of the systems thoroughly and continue working on it because they make both obvious and hard-to-detect mistakes.

This means that generative AI today is only a good tool in areas where the amount of time spent on corrections and adjustments does not exceed the value the tools give you. In some areas, the problems are worse than others. The systems are, for example, weak in areas where timeliness and factualness are crucial. Here, both strong prompt engineering skills and thorough checking of the output are required to use the systems professionally.

“Prompt engineering” is becoming one of the most hyped concepts in generative AI because it currently requires strong skills in prompt engineering to get good results when working with the systems. At the same time, the need for prompt engineering may be an indication that interaction with the generative models is not sufficiently intuitive yet. Prompt engineering may therefore be a transitional phenomenon on the way to more mature and user-friendly generative models.

At the same time, generative AI systems will improve in the coming years. Both OpenAI, which developed ChatGPT, and many other companies are working intensely to improve generative models, and as early as 2024, we will see new systems that outperform the best from 2023.