The Underground Basket-Weaving Society of Aus.

1. make a syllabus for yourself - books, media, places, recipes

2. complete 40% of it

- it’s fun to be awake & in an upright position

- consciousness is a gift

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.

things fall apart and there’s anarchy everywhere; falcons fly freely and can’t be compelled.

Dimming the tide there is blood in the water and everywhere customs of innocence drowned.

Held by the worst is a fervent intensity; lacking conviction the best run aground.

Surely there must be some kind of apocalypse; surely epiphany must be at hand!

Visions unbidden from Spiritus Mundi of something awakening out on the sand:

Head of a man with a leonine frame and a pitiless gaze that’s as blank as the sun;

sleep became nightmare by passage of centuries long before eons of slumber were done.

Higgledy piggledy, beast of antiquity, imminent, just as the prophecies warn,

[Q: Why did you decide to do this experiment?

A: Snake behavior has been generally neglected as a field of research, especially in Brazil. And most studies don’t examine what factors make them want to bite. If you study malaria, you can research the parasite that causes the disease—but if you don’t study the mosquito that carries it, you will never solve the problem. Up until now, the popular wisdom was that the jararaca would only attack if you touched it or stepped on it. But that was not what we found.

Q: Why did you need to be the victim?

A: The best way to do this research is to put snakes and a human together. In this case, the human was me. We put the snakes inside a ring on the floor of our lab until they got used to it, then I stepped in wearing special protective boots. I stepped close to the snake and also lightly on top of it. I didn’t put my whole weight on my foot, so I did not hurt the snakes. I tested 116 animals and stepped 30 times on every animal, totaling 40,480 steps.]