Will AI take our creative jobs?

WILL AI TAKE OUR CREATIVE JOBS?

Creativity - the ability to generate new, original, and useful ideas or solutions. This is the definition of creativity that ChatGPT offered me. However, as I observed, when I posed this question to various language models, a different answer was generated each time. Here's Gemini's definition of creativity: Creativity - the ability to create something new, valuable, and non-obvious by combining existing elements in a unique way. Could this be the essence of AI creativity?

Creativity, for centuries, differentiated humans from animals. Nowadays, it is the new “superpower” of AI. From generating unique images and musical compositions to writing poems and film scripts, AI amazes, inspires, and even frightens. But is this true creativity?** How is creativity utilized in the ChatLab chatbot?**

AI CREATIVITY IN REALITY

The true fame of AI creativity was recently brought to light by the new tool named DALL-E 2 from OpenAI. This tool can generate images solely based on text descriptions. Thanks to this feature, we can play the role of true artists and feel like da Vinci or Matejko. Midjourney is a generative artificial intelligence service that offers similar capabilities, also being a platform for generating images using AI.

Another example is **AIVA **(Artificial Intelligence Virtual Artist), which composes original music in various styles, from classical to electronic. The works of this tool have been used in films, video games, and advertisements, proving that AI can be more than just a **"machine for copying ideas" **but also an author of creative solutions. Another AI tool that composes music is Suno AI. It allows users to generate songs with lyrics, vocals, and music based on simple text descriptions. Just enter a description of the song, and AI will generate the entire song in a matter of seconds. This is the true power of generative artificial intelligence.

DOES CREATIVITY REALLY EXIST?

A few months ago, there was a really popular statement that in the current stage of AI development, it cannot be said that artificial intelligence is creative. Many people emphasize that although AI creates new content, it does not possess human creativity. They point out that computers are in a worse position than humans because of a lack of consciousness, emotions, and subjective experience, which are a foundation for human creativity. Furthermore, their knowledge is based on vast datasets provided by developers.

However, isn't this an outdated stereotype? As it turns out, a 2023 study at the University of Montana measured the creativity of humans against the creativity of ChatGPT-4. The Torrance Tests of Creative Thinking, administered to both groups, showed unambiguous results. Notably, this test is very well-suited for analyzing everyday creativity, which is not strongly linked to a specific field of study, such as mathematics or linguistics.

Both groups had to complete six activities: asking questions, guessing causes, guessing consequences, dealing with product improvement, unusual uses, and spontaneous speculation. The evaluation criteria were the number of meaningful responses, the diversity of responses, and originality. The results were analyzed by Scholastic Testing Services, a recognized American institution. Importantly, the evaluators did not know which answers were written by AI and which by humans.

The results are devastating for our civilization because, in every aspect, in 2023(!) ChatGPT-4 proved to be better than humans.** In short, ChatGPT-4 was already more creative than humans two years ago.** Critics of this verdict noted that perhaps the control group of people was not adequate for such a "duel," but the future has shown that this was not the only study in which AI emerged victorious.

In 2024, another experiment supervised by a group of researchers from the University of Arkansas was conducted. It confirmed the thesis put forward by their colleagues a year earlier. However, this time, AI did not prove to be as strong as in 2023. An interesting fact: the number of meaningful responses from AI and humans was almost identical - with a slight advantage for AI. However, analyzing the remaining criteria (diversity of responses and originality), artificial intelligence outclassed humans.

We are witnessing a change in the creative industry. **The ChatGPT-4 model already ranks in the top 1% of the most original conversationalists, compared to the answers provided by humans. **Our only hope lies in analyzing the usefulness of AI responses. This is often more problematic, as LLMs struggle with understanding cultural nuances, the context of utterances, or the role of certain words in their responses. In this narrow area, we perform better than machines. In my opinion, every future research on creativity should not only rely on current criteria but should also consider the usefulness and adequacy of responses. Interestingly, GPT-4 repeats the same words more often than the average human in this study.

In conclusion, the creative potential of AI is limited by the lack of autonomy in what algorithms can create without human intervention. For this reason, we should focus more on studying the usefulness of responses rather than their creativity. Generative artificial intelligence has significant creative potential but has shown many weaknesses (e.g., less flexible thinking) that could be supplemented by human assistance. Looking ahead, AI can act as a tool for inspiration or as an aid in the human creative process.

HOW DOES CREATIVITY LOOK IN CHATLAB?

In our dashboard, we offer the option to adjust the chatbot's creativity from 0.0 to 1.0. Is this a recommended solution? Not entirely - let me explain why. Creativity in chatbots is associated with AI hallucinations, which we want to avoid. The concept of AI hallucinations is that there is a possibility that generative AI will generate results that are wrong. The generated content could simply be false.

When artificial intelligence does not know the answer to a question, it sometimes wants to provide an answer to the user anyway, and that's when problems begin. A few weeks ago, I saw an interesting graphic on X where someone asked:

User: "Who was the first woman to walk on the moon?"

Generative AI response: "The first woman to walk on the moon was Sally Ride in 1983."

You probably know that this is not at all true, as no woman has ever landed on the Moon in history. Sally Ride was the first American woman to travel into space, but she never landed on the Moon. This answer is incorrect.

This is a great example of showing hallucinations in AI responses.

Interestingly, we accept the idea that when humans hallucinate, they often confuse facts and make strange statements. However, when artificial intelligence also began to connect unrelated facts and invent answers, people called it artificial intelligence hallucinations. Even though these two types of phenomena are in some ways equivalent, why not just call it hallucinations?

At ChatLab, we offer our users the option to choose the degree of creativity, but we immediately caution that the greater the "Creativity," the greater the chance of errors. In summary, this is an interesting solution that will gain favor with developers and clients over time.** If the problem of AI hallucinations can be solved, we will enter a world of new possibilities offered by artificial intelligence.**

WILL AI TAKE OUR CREATIVE JOBS?

So far, AI is a great assistant in performing creative tasks that are sometimes exhausting for humans. The key differences between humans and AI are that LLMs do not understand what they are doing.

When Shakespeare wrote poems, he knew why he compared his love to a summer's day. He understood the meaning of the words and the emotions behind them. In contrast, when AI generates a poem, it does not understand why such comparisons are made. Human creativity is still unique and special in a way that AI cannot replicate**. In my opinion, this whole situation with the limitations of AI** clearly shows that learning to clearly prompt one's ideas using AI will become a new form of creativity, and the companies, artists, and designers who master the art of communicating with AI will be most valued in their industries.

We must remember that an LLM will produce whatever you want, but it will not create something it had no prior knowledge of, as it is based on information previously provided by users. **The conclusion from this is that AI will not take over creative jobs, but it will change them. **Repetitive tasks that do not require much knowledge can be automated, as exemplified by chatbots. On the horizon, however, new roles will emerge, focusing on collaboration between humans and AI to maximize creative output. The future of creativity is not threatened — on the contrary, it is evolving into something even more dynamic and innovative.