Skip to content
All posts

How ChatGPT Fumbled Its Own Challenge. Are You Better Than AI?


Creating and Solving Riddles–An Inherently Human Task?


In a fascinating experiment, we decided to test the limits of artificial intelligence by giving ChatGPT, an advanced language model, the seemingly simple task of creating and solving a riddle. The riddle is as follows:


I can be cracked, made, told, and played,
I can bring laughter, tears, and aid.
I'm a vessel of stories, both old and new,
With words and pages, I come into view.
What am I?


What transpired left us astonished and puzzled. Despite its immense capabilities and vast knowledge, ChatGPT encountered unexpected difficulties, struggling to solve the very riddle it had crafted. We repeated the experiment multiple times, and ChatGPT failed to provide the correct answer to its own riddles roughly two-thirds of the time. This article delves into the obscure nature of the riddle, the potential reasons behind ChatGPT's failure, and what it reveals about the complexities of AI systems. 


DALL·E 2023-05-25 14.14.29DALL·E 2023-05-25 14.09.10 - ChatGPT solving a riddle in van goghs style

Dall E's take on "ChatGPT solving a riddle in VanGogh style"


Unraveling the Riddle:

The riddle presented by ChatGPT immediately captivated our attention. We were eager to see if the AI could swiftly deduce the answer based on its extensive knowledge base. However, to our surprise, ChatGPT struggled to solve the riddle it had crafted.


Possible Explanations:

Semantic Ambiguity: The riddle contained several elements that were open to interpretation. For instance, the concept of "vessel" and "view" could be perceived metaphorically or as abstract representations. The inherent ambiguity might have thrown ChatGPT off track.

Contextual Limitations: While ChatGPT possesses an impressive database of information, it is still bound by its training data and knowledge cutoff of September 2021. If the riddle contained references or context beyond this timeframe, ChatGPT might have been at a disadvantage, hindering its ability to solve it.

Creative Thinking Challenges: Riddles often require an imaginative and lateral thinking approach to solve. While ChatGPT is adept at processing and generating language, its creativity and ability to think outside the box may be limited compared to human counterparts. The abstract nature of the riddle may have exceeded its capabilities in this regard.

Implications and Reflections: ChatGPT's inability to solve the riddle it had created offers insights into the limitations of current AI systems. Despite its immense computational power and access to vast amounts of information, ChatGPT struggled with an abstract and ambiguous puzzle. This demonstrates that there are still challenges to overcome in AI's ability to comprehend complex language nuances, metaphors, and creative thinking.

Moreover, this experiment highlights the importance of continual improvement and evolution in AI systems. While ChatGPT may have stumbled in this particular instance, it is crucial to recognize that it has the potential to adapt and grow through further training and refinement.

The enigmatic riddle presented to ChatGPT led to unexpected results, leaving the AI model stumped by its own creation. The obscure nature of the riddle, coupled with the complexities of language interpretation and creative thinking, proved to be formidable challenges for the AI. This experiment serves as a reminder that despite the tremendous advancements in AI technology, there are still frontiers to explore and obstacles to overcome on the path towards human-like intelligence.


The answer to that riddle, by the way, is a joke. (ChatGPT's answers include: a book, a smartphone, and eventually, a joke).