Artificial Intelligence (AI) is changing the world in many ways. As heavily reported, many workers, such as those in content creation, data entry, or customer service fields, are at risk of losing their jobs. In fact, according to a study by the McKinsey Global Institute, 14% (or about 1/7) of the global workforce may lose their jobs and change careers by 2030 due to automation. There are several significant reasons behind such a dramatic shift. One is that with fewer workers, there will theoretically be fewer costs, which will improve organizations’ bottom lines. Another is that with an automated system in place, there would theoretically be fewer mistakes made, as well as a higher level of efficiency.
However, what may make sense in theory may not hold true practically. For example, in May 2025 a writer working on behalf of King Features Syndicate, a division of the Hearst Corporation, wrote an article that detailed a recommended summer reading list. The article was picked up by several newspapers, including the Chicago Sun-Times and Philadelphia Inquirer. The only issue was, as it turned out, some, if not all, of the books on the list did not exist. The writer had used AI to help create the piece, and the software had invented book titles and attributed them to real authors.
While the above example illustrates both the limitations of AI and the importance of checking one’s work, the use of the technology in the above scenario is appropriate; obviously, a significant capability of AI is to generate ideas. Platforms such as ChatGPT can create long form responses to essentially any prompt within a few seconds. These answers are generally detailed, well-organized articles that prove informative to those that read them. But, given the fact that ChatGPT and other platforms like it use data that is entirely available to and compiled by humans (or at the very least directed by humans), it seems that AI might have some of the same limitations in its ability to create new ideas that humans experience.
To obtain a sense of whether or not our hypothesis was correct, we asked ChatGPT several questions to assess its capabilities:
Question 1: Will AI always be able to come up with new ideas?
“
Short answer: Not exactly always, and not exactly new in the way humans mean it—but it can stay fresh, to a point.
Here’s why:
1. AI’s creativity is based on patterns
AI generates ideas by identifying patterns across massive amounts of data—books, art, code, speech, etc.—and remixing them in new ways. That can feel very creative and often is in practical terms.
But:
AI doesn’t have intuition or lived experience.
It doesn’t want anything or wrestle with meaning the way humans do.
It lacks what you’d call a “why” behind its ideas.
So even though it might keep combining patterns forever, whether those combinations feel fresh or stale depends on how diverse and updated its training data is—and what you’re asking it to do.
2. It depends on the data and updates
AI can start to feel stale if:
It stops learning from new input (e.g., art, culture, user behavior).
The prompts it receives are repetitive or lack depth.
It over-optimizes for what it thinks is useful or expected.
Freshness requires:
Diverse, evolving training data.
Exposure to edge cases, culture shifts, unexpected problems.
Human guidance—people pushing the AI to stretch.
3. What about self-refreshing AI?
Future models might be able to:
Learn continuously (online learning).
Self-critique and evolve (with help from reinforcement or human feedback).
Use tools (like experiments or the internet) to extend their creativity.
But even then, they’ll reflect the limits of what they’ve been allowed or designed to explore. If the world stops asking hard or new questions, AI won’t invent them on its own.
So will it get stale?
It can, yes—if:
It’s locked into old training data.
It’s not prompted in diverse or challenging ways.
Its creators don’t prioritize novelty or weirdness.
But with the right input and systems, AI can keep surprising us for a long time—just like a good jazz musician who’s always listening, learning, and riffing.
”
Question 2: Would AI be able to come up with different solutions to the same problem?
“
Yes, AI can come up with different solutions to the same problem—especially if it’s designed to be creative, adaptive, or exploratory in its approach.
Here’s how and why that happens:
1. Multiple Valid Paths
Many problems (especially in areas like design, writing, strategy, and engineering) have more than one correct or effective solution. AI models can generate a range of these depending on:
How the problem is framed
What constraints or goals are emphasized
Random variations in generation
2. Stochastic Nature
AI models like ChatGPT use probabilistic methods to generate outputs. That means even when given the same input multiple times, the model can produce different, but still relevant or correct, responses—unless it’s set to be deterministic (with a low temperature setting).
3. Different Perspectives
AI can also emulate different perspectives, styles, or methodologies:
A legal issue might be approached from a business, ethical, or regulatory perspective.
A writing prompt might be handled in a poetic, journalistic, or academic tone.
A math or logic puzzle might have an intuitive explanation or a formal proof.
4. Prompt Engineering Matters
Changing even slightly how a question is asked can cause the AI to:
Prioritize different information
Choose different reasoning paths
Offer alternative types of answers (step-by-step, summary, analogy, etc.)
Example
Problem: “How can a company reduce customer churn?”
One answer: Improve onboarding and customer support.
A second answer: Offer loyalty programs and discounts.
A third answer: Use predictive analytics to intervene early.
Each of these could be valid, and AI can suggest all or focus on one depending on the context.
”
The responses are intriguing. They confirm that humans drive the efficacy of AI and are largely responsible for its ability to continue to produce relevant, new information. It also confirms that AI is devoid of important parts of being human, such as intuition or lived experiences, which would improve its ability to develop fresh ideas. But it also derives its capabilities from humans, who obviously do have lived experiences, which provides it with at least some of that important knowledge to develop new ideas. It also processes massive amounts of information, far more than a human could, in a very short amount of time, which provides tremendous value. Additionally, AI has the potential for future improvement, such as continuous online learning, which could increase its worth even more.
So, maybe the most appropriate way to think about AI is not so much that it can generate unique ideas of its own, but that it represents a conglomeration of ideas from the entire global population. It is the combination of this, as well as an algorithm that outlines the use of that information, that form a tool to complement existing human work and functions. The result is that ideas can be accessed and exchanged more quickly, creating a more interconnected world. It seems that, as long as humanity does not run out of new ideas and have new experiences, AI will continue to bring all different types of ideas together from every corner of the globe.
Artificial Intelligence – A Work in Progress
