Hardly any technology has been hyped in recent years as much as ChatGPT, the new chatbot from OpenAI. An endless number of social media posts are filled with application tips. At the same time, some authors conjure up scenarios of threats or foresee the decline of educational institutions or even the entire job market. But what is really behind the hype?

What is ChatGPT anyway?

The foundation for ChatGPT is actually nothing new. ChatGPT is based on GPT3, a so-called Large Language Model ("LLM"), which was already released in 2022.

As ChatGPT itself states, a "Large Language Model" (LLM) is an artificial intelligence algorithm that is based on machine learning and is trained to understand, process, and generate human language. These algorithms are often used for tasks such as translation, text summarization, question-answer systems, and chatbots. LLMs are capable of analyzing large volumes of language data and recognizing patterns to generate human-like responses to questions or tasks.

There are already several LLMs, and even Yann LeCun, the chief AI scientist at Meta (the company behind Facebook, Instagram, and WhatsApp), says ChatGPT is "not particularly innovative" and "nothing revolutionary." And in principle, he is right. However, what is truly new in the case of ChatGPT is the extremely simple way in which OpenAI has made such a powerful tool accessible to ordinary users.

First of all, ChatGPT is currently freely accessible to everyone. You have to register, but that is no longer an obstacle for most users these days. After a short waiting period, you will be activated and can access the bot.

Then, the interface of ChatGPT is incredibly simple - you just enter your question or instruction (English: "Prompt") and receive an answer. As long as you formulate your question concretely and provide some context if necessary, you typically get a meaningful response - sometimes even an incredibly good and detailed one. In some cases, it feels like magic. And that's precisely where the success of the tool lies.

So fast an adoption as no tool before

Only two months after the launch at the end of November 2022, the tool already had 100 million monthly active users by January 2023, according to data from Similarweb (Source). These users were primarily acquired through word of mouth and not through advertising campaigns - simply because enthusiastic users talked about it.

For comparison: Instagram took two and a half years to reach 100 million. TikTok managed it in nine months.

What applications are there?

In principle, you can ask ChatGPT anything you want to know, as long as it is limited to knowledge up to and including 2021 – the training data extends only to this point.

For example, ChatGPT can create recipes, generate ideas for LinkedIn posts, write programming code, or explain complex topics like quantum computing. The whole thing is natively in English, but it also works in German. Something that not everyone knows, and what is relatively new from an AI usage perspective, is that you can also ask follow-up questions.

To demonstrate these points, we asked ChatGPT to develop a marketing strategy for a fictional podcast:

Not a bad start, although still quite generic. I then asked for a bit more detail and practical help on one of the points:

You could have each point further elaborated, create additional examples, or even write a complete blog post. This depth and type of interaction with the AI is new, practically helpful, and sometimes truly impressive.

ChatGPT's answers are rarely innovative. However, they can be very helpful to start with a task and receive possible solutions and inspiration.

Especially with topics that have already been solved somewhere in the world and for which there are blog articles or discussions on the internet, ChatGPT is already often a quicker help than Google.

Are there risks?

ChatGPT is an impressive technology. And yet, there are potential risks in usage, especially in a business context.

On one hand, ChatGPT always responds very convincingly, but not always factually correctly. Occasionally, a book is invented, for example, that the author did not actually write, or suggestions for action are made that do not make real sense in context. Therefore, the answers should always be double-checked.

Furthermore, ChatGPT can also be biased and, for example, discriminate. There is an active discussion about the extent to which technology should have a political orientation or remain as neutral as possible, and whether this should be actively controlled by OpenAI. A "correct" approach certainly still needs to be found.

Additionally, there may be ethical and legal risks, as the AI was trained on very large datasets based on content created by other people, and the results are now being used for business purposes.

But ultimately, it is like with any revolutionary technology: once it becomes available and promises significant benefits, it will also be utilized. Everyone should therefore proceed with caution and consider how they can use it positively and ethically to create, for example, more creative solutions, provide better customer service, or simply save time that can be used productively elsewhere.

What comes next?

There is no question that ChatGPT has made a huge impact. Its usage in many office and creative situations will only continue to grow, and more and more tricks for effective interaction with the tool will certainly be found (Keyword: "Prompt Engineering").

In addition to direct usage, many other tools are emerging that are based on ChatGPT or GPT3 and offer industry-specific solutions – whether it is email programs, chatbots, or all aspects of content creation.

Even Microsoft is working on implementing OpenAI's technology in its Office Suite. And Microsoft has now also introduced a competitive solution to ChatGPT with Bing Chat. When you engage with it, it has a slightly different character and sometimes even offers more impressive results.

So, a lot is happening, and these are just initial approaches that do not even consider other content types and interaction forms (ChatGPT focuses on text-to-text interaction, but there are certainly also tools that allow for speech-to-text, text-to-image, and many other forms).

Overall, it remains to be noted that while it is indeed hype, there is also substance behind it, as real added value is being created for many users. And the technology is advancing rapidly - so it remains exciting!