ChatGPT Failed a Basic Observation Puzzle – Needs Improvement

I am a regular user of ChatGPT and recently tested it with a simple observation-based reasoning puzzle. However, to my surprise, the AI failed to notice an obvious pattern and gave a completely incorrect answer with full confidence.

The Puzzle:

Find the odd one out:
2, 28, 48, 58, 128

Correct Answer & Explanation:

By simple observation, all numbers except one contain the digit ‘8’:

28 → Has ‘8’

48 → Has ‘8’

58 → Has ‘8’

128 → Has ‘8’

2 → Does NOT have ‘8’

Thus, the correct answer is 2, as it is the only number without the digit ‘8’.

Issues with ChatGPT’s Response:

  1. It initially gave a wrong answer (128) based on unnecessary mathematical logic.

  2. It later incorrectly claimed that “58 does NOT have 8”—a direct observation mistake.

  3. It was overconfident in a wrong answer and didn’t self-correct.

Suggestions for Improvement:

Better attention to basic visual details in reasoning questions.

Self-verification before providing final answers to avoid simple mistakes.

Balancing logical & observational thinking to ensure better accuracy in such cases.

I believe such improvements would make ChatGPT more reliable, especially in basic reasoning tests where users expect accurate responses.

Looking forward to your response.