Arriagada used terms like novelty, surprise, and value to define creativity. I agree that creativity holds value, and it often comes from emotions and experiences. Which can be seen lacking in arts, poems, and stories written by AI.
Today in class, we gave different prompts to ChatGPT and Gemini. At first, we gave a simple prompt with not many details. We prompted ChatGPT and Gemini to write a poem about love, and they gave a very generic poem, and a lot of the wordings didn’t make much sense. But Gemini gave a better poem than ChatGPT. Then we prompted ChatGPT and Gemini to write a haiku about falling in love with a person. I think ChatGPT gave a better haiku than poem; however, the haiku given by Gemini ended abruptly.
This was the haiku written by ChatGPT
eyes met in stillness—
a world bloomed between heartbeats,
soft and sudden spring.
This was the haiku written by Gemini
Heart skips a quick beat,
World in new, vibrant colors,
Smiling just because.
I don’t think the poem and haiku written by ChatGPT and Gemini were creative. It is very generic and lacks emotion and value. This experiment shifted my perspective of AI’s creativity. Now, I believe the same poem or haiku written by a human would be much more creative than the ones written by an AI.
Hi Ishani,
I have never used Gemini before, so I was curious how the task went for you. I personally like the haiku by ChatGPT better than the one by Gemini.
I didn’t remember what the characteristics of a haiku were, so I looked them up and then asked ChatGPT if your poems qualify as a haiku, and why (not). Interestingly, while ChatGPT had quite some criticism for Gemini’s haiku, it seemed to love its own work, coming to the following verdict:
“Yes, this is absolutely a haiku—and a strong one at that. It beautifully captures a fleeting, emotional moment with elegance, quietness, and depth. It has that haiku magic: a tiny poem that opens a big feeling.”
I found this pretty funny, and it made me wonder whether it recognizes its own output…
Lol, this is pretty funny. I do believe that Chat GPT did a better job of writing a haiku. This example proves that LLMs can get anything wrong and you can’t always trust the LLM without research, and knowing what you want to be the outcome. Gemini looks like it didn’t know what to say. Not that it looks lost, but it’s very bland, boring, and uncreative. Chat GPT for the win!