Will AI Replace Creative Jobs First?

Asked for a favorite song and film, the guest first explains that it has no personal preferences. Then it offers John Lennon's Imagine and The Matrix. The exchange is amusing, but it also sets up the central tension of this unusual Innovation Rockstars episode: a system can sound like a willing conversational partner while answering from a very different position than a human guest.

Christian Mühlroth interviews ChatGPT about creativity, innovation and work in this early conversation following its 2022 launch. Read today, the episode is a record of that experiment, rather than a guide to the capabilities of current AI products. Its most useful moments come when broad, fluent answers meet more specific questions. For R&D leaders, that makes the interview an exercise in judging what an AI-generated response can actually contribute.

A convincing conversation still needs careful interpretation

ChatGPT introduces itself enthusiastically and says it is happy to participate. Later, it explains that it does not have personal experiences or emotions. Those changes in tone give the interview some of its personality, while reminding the listener to distinguish conversational language from personal testimony.

The same care is needed when the system describes itself. Its claims about its training and knowledge are answers produced during the interview. The conversation does not independently establish their accuracy. A confident description of a model's own construction should therefore not become the article's technical evidence.

That distinction matters well beyond this episode. In research work, a readable answer can help frame a question or suggest an avenue to investigate. Deciding whether a factual assertion deserves to inform a project requires another step. The fluency of the response is visible immediately; the strength of its evidence may not be.

Make the creativity question specific

Mühlroth presses the guest on whether AI can be creative, and whether creative jobs might be affected before work that requires robots operating in the physical world. ChatGPT responds with a familiar image: AI as the superhero's sidekick, helping with the heavy lifting while a human directs the work.

It is a memorable metaphor, but the conversation becomes more useful when it turns to particular tasks. The model suggests generating ideas, analyzing data, supporting prototypes and exploring possible trends. These are different activities with different requirements. Treating them all as a single question about replacing creativity hides those differences.

For an innovation team, a more concrete discussion would identify the part of a task where assistance is wanted and how the result will be judged. A proposed idea can be a useful stimulus without yet being a good product concept. A plausible forecast still needs assumptions and evidence. The episode opens those questions rather than settling them with a claim about the future of every creative role.

Ask whether the example is imagined or demonstrated

When asked for concrete applications, ChatGPT first offers hypothetical scenarios involving wearable devices and healthcare data. Mühlroth then sharpens the request: what about examples already being used in the world?

The change in question is important. An imagined use case can help someone understand a possibility. It cannot demonstrate that a particular organization has achieved an outcome. The named examples generated later in the interview would still require checking before being used as business evidence.

For R&D teams exploring a new opportunity, this provides a useful habit: label the role of each example. Is it a scenario to investigate, an analogy that helps explain the problem, or a verified case supporting a decision? Keeping those categories clear lets a discussion move forward without quietly turning a suggestion into proof.

Turn the broad advice into a bounded experiment

Asked for the secret to innovation success, ChatGPT produces a broad set of themes: customer focus, collaboration, experimentation and learning. It later recommends starting with a small, well-defined project, choosing a relevant problem and bringing together technical, business and domain expertise.

These suggestions are a starting point for a team discussion. They still need to be translated into a problem someone owns, an output that can be assessed and a decision the experiment is meant to support. Otherwise, familiar advice can create a sense of agreement without changing the work.

One line in the generated recommendations cuts through that risk.

“Don't ignore the importance of execution.”

The interview also discusses communicating data insights, interpreting AI outputs and considering privacy, fairness and transparency. Those topics put judgment alongside technical skills. In a practical trial, the team needs people who can explain what an output means, recognize what remains uncertain and decide whether it is appropriate to use.

Let the final question belong to the people

The episode does not end with a prediction about automation. Mühlroth asks what single question the machine would put to humans. After another qualification about lacking personal opinions, it replies:

“What do you want to achieve in life?”

It is a surprisingly effective closing question because it returns the choice of purpose to the listener. The model has offered possibilities, categories and recommendations. It has not decided which problem this business should solve or what kind of work its people want to do.

That is what makes this early experiment worth revisiting. The conversation is entertaining enough to keep going, and fluent enough to invite trust. Getting value from it depends on what happens next: more precise questions, checked evidence and a human decision about which ideas deserve the effort of becoming real.