Archive · 20 Aug 2025
ChatGPT-5 Learns Honesty, Says 'I Don't Know,' Praised by Musk
Imagine a super-smart computer program, ChatGPT-5, that can answer your questions.
Unlike earlier versions, if it doesn't know something, it now says "I don't know" instead of making up an answer.
This is seen as a good thing, like the AI being honest!
Even Elon Musk thinks it's impressive.
However, the people who made ChatGPT-5 are still cautioning you not to trust it completely, saying that it's best to use it as a second opinion, not as the only source of information.
ChatGPT-5 now admits it does not know the answer to some questions.
The new feature is considered a step forward for AI accountability.
Elon Musk praised ChatGPT-5's ability to avoid incorrect answers.
OpenAI's head of ChatGPT warns against over-relying on the AI.
The new model reduces incorrect answers to approximately 10%.
- Who
- ChatGPT-5 and its developers at OpenAI; Elon Musk commented on the development.
- What
- ChatGPT-5 now admits it doesn't know the answer to certain questions.
- When
- The new feature is implemented in the GPT-5 model.
- Where
- N/A
- Why
- To avoid giving incorrect information and to improve the reliability of the AI.
Users and Developers
OpenAI
Reliability
Users and Developers
The new feature is an improvement, making the AI more reliable.
OpenAI
The AI should not be solely relied upon as it still can be inaccurate. Always double-check the information.
- Model
- GPT-5
- Key Change
- Admits "I don't know" to some questions
- Accuracy Improvement
- Reduced incorrect answers to 10%
- Elon Musk's Reaction
- Called the feature "impressive"