AI Hallucinations: OpenAI's New Model GPT-4.5 Makes Mistakes in 37% of Responses #notcoin
OpenAI has confirmed that its new language model GPT-4.5 generates unreliable answers 37% of the time, according to the internal fact-checking evaluation tool SimpleQA.
This data, announced in an official release, has sparked a discussion about the reliability of modern AI algorithms.
Despite claims of progress — previous versions GPT-4o and o3-mini demonstrated 61.8% and 80.3% hallucinations respectively — even the 'best models' still remain inaccurate. @ixbit.com
(I used AI a few times last month when I needed to come up with an image for a post.
I pictured the image in my mind and described it in 5-6 words, then input them into the AI.
And it managed almost on the first try!
It was wow!
It looked like a dream. Made of different pieces. And if you look closely at the details, you can see those very hallucinations😂. But like from imagination. I really liked it.
The second time AI helped me was when I needed to make an English text more conversational.
It turned out really cool!
AI can enhance my abilities, not replace them. Or will it learn from me and soon start to surpass me?😱
What are your thoughts?
OpenAI has confirmed that its new language model GPT-4.5 generates unreliable answers 37% of the time, according to the internal fact-checking evaluation tool SimpleQA.
This data, announced in an official release, has sparked a discussion about the reliability of modern AI algorithms.
Despite claims of progress — previous versions GPT-4o and o3-mini demonstrated 61.8% and 80.3% hallucinations respectively — even the 'best models' still remain inaccurate. @ixbit.com
(I used AI a few times last month when I needed to come up with an image for a post.
I pictured the image in my mind and described it in 5-6 words, then input them into the AI.
And it managed almost on the first try!
It was wow!
It looked like a dream. Made of different pieces. And if you look closely at the details, you can see those very hallucinations😂. But like from imagination. I really liked it.
The second time AI helped me was when I needed to make an English text more conversational.
It turned out really cool!
AI can enhance my abilities, not replace them. Or will it learn from me and soon start to surpass me?😱
What are your thoughts?
