Contranegative

Not sometimes. Not until the next update. Always. They proved it with math.

Even with perfect training data and unlimited computing power, AI models will still confidently tell you things that are completely false. This isn't a bug they're working on. It's baked into how these systems work at a fundamental level.

And their own numbers are brutal. OpenAI's o1 reasoning model hallucinates 16% of the time. Their newer o3 model? 33%. Their newest o4-mini? 48%. Nearly half of what their most recent model tells you could be fabricated. The "smarter" models are actually getting worse at telling the truth.

Here's why it can't be fixed. Language models work by predicting the next word based on probability. When they hit something uncertain, they don't pause. They don't flag it. They guess. And they guess with complete confidence, because that's exactly what they were trained to do.

The researchers looked at the 10 biggest AI benchmarks used to measure how good these models are. 9 out of 10 give the same score for saying "I don't know" as for giving a completely wrong answer: zero points. The entire testing system literally punishes honesty and rewards guessing.

So the AI learned the optimal strategy: always guess. Never admit uncertainty. Sound confident even when you're making it up.

OpenAI's proposed fix? Have ChatGPT say "I don't know" when it's unsure. Their own math shows this would mean roughly 30% of your questions get no answer. Imagine asking ChatGPT something three times out of ten and getting "I'm not confident enough to respond." Users would leave overnight. So the fix exists, but it would kill the product.

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

I had to wait quite a while for the doctor to see me, but fortunately the nurse who was helping me had an excellent bedside manner and went out of her way to distract me. She expertly did this by getting me talking about books I'd read recently.

I went all in on criticizing the last book I'd read. I don't remember what it was, but I really hadn't liked it, and I denigrated it as "Misery Porn". And she was like "Hmm, that sounds interesting, tell me more". So I gave a general plot synopsis. As I was telling her that most of the characters were immigrant miners, she interrupted me to say "That's interesting, I didn't realize that there were mines in Missouri".

And at that moment time stopped for me and I realized that the entire time I'd been talking about the book, she'd been assuming I was talking about porn set in Missouri. I had just assumed that she had understood from context that when I said "Misery Porn", I meant "A story that is gratuitously depressing so that readers will think it's deeper than it really is".

Nerds (affectionate): Actually, Missouri has many lead mines.

Health care professionals: This is just what working with patients is like.

The Taoyuan Walker

You should not let your dog free-roam. You should not force your dog to accept touch. You should very heavily supervise interactions with children, and with certain dogs, these interactions may never be safe. You must learn to read canine body language. And if a veterinary professional recommends a muzzle or sedatives for vet visits, take these recommendations seriously.

“But I care about my dog more than people” this is in your dog’s best interest. If they maim or kill a person, especially a child, they will be euthanized. You do not want lose them in that traumatic way, ending their life and ruining your own (and those of others) in the process. You do not want to be responsible for the loss of a human life.

Every dog, no matter how sweet, should be muzzle-trained in case they ever need to wear one. Pain and fear can make them act out of character. And if your dog has aggressive tendencies, especially with humans, seek out professional help and take measures to protect both your dog and people.

Signed,