🚨 BREAKING: OpenAI has published an article proving that ChatGPT will always make things up.

Not sometimes. Not until the next update. Always. They proved it mathematically.

Even with perfect training data and unlimited computing power, AI models will always confidently tell you things that are completely false. It’s not a bug they’re working on. This explains how these systems fundamentally operate.

And their own numbers are brutal. OpenAI’s reasoning model o1 hallucinates in 16% of cases. Their new model O3? 33 percent. Their new o4-mini? 48 percent. Nearly half of the information provided by their latest model could be fabricated. "Smarter" models are actually getting worse at telling the truth.

Here’s why this can’t be fixed. Language models work by predicting the next word based on probability. When they encounter an uncertain situation, they don’t stop. They don’t signal it. They guess. And they speculate with full confidence because that’s exactly what they were trained to do.

Researchers examined the top 10 AI criteria used to measure these models’ quality. 9 out of 10 give the same score for saying "I don’t know" as for giving a completely false answer: zero points. The entire testing system literally punishes honesty and rewards conjecture.

AI has therefore learned the optimal strategy: always guess. Never admit uncertainty. It appears confident even when inventing.

What is OpenAI’s proposed solution? Ask ChatGPT to say "I don’t know" when it’s unsure. Their own calculations show that this would mean about 30% of your questions would go unanswered. Imagine asking ChatGPT three times out of ten and receiving "I’m not confident enough to answer." Users would leave overnight. The fix exists, but it would kill the product.

This isn’t just an OpenAI problem. DeepMind and Tsinghua University independently reached the same conclusion. Three of the world’s top AI labs, working separately, all agree: it’s permanent.

Every time ChatGPT gives you an answer, ask yourself: is this real or just a confident guess?

Here’s Come-from-Beyond’s 👇 response:
They are starting to understand why

#QUBIC #Aigarth .* has made the ability to say "I don’t know" a core concern

$QUBIC