Untitled

"Demagolka" (n., Mando'a) [ˌdɛməˈ ɡoʊl kə]

A real-life monster, a war criminal; one who commits atrocities, especially against children or the vulnerable.

Earlier today I made a reference to “Backstroke of the West”, and when I looked it up I saw that the original post was made in 2005, which means that a good chunk of people on here might be too young to remember it. So consider this post a public service.  The following was originally posted here: 

6.07.2005

episode iii, the backstroke of the west

i saw revenge of the sith last weekend at a local theater with my friend joe who was in town on business. it was much better than the first two movies and a fitting end (err.. middle) to the star wars saga. the next day i was walking past my friendly dvd salesperson and decided to check out revenge of the sith. i was assured the quality was good and for 7rmb why not give it a shot. aside from the counters on the top of the screen and a distorted perspective it was ok- not high quality but watchable. the captions were a hilarious surprise- a direct english translation of the chinese interpretation of what the script was saying. it varied from being somewhat close to the script to being ‘far far away’….

amazingly enough, the beginning scroll is mistranslated even though the words are right there on the screen.

star war (just one)

'the backstroke of the west’ is the english translation of the chinese title.

anakin: “this is where the fun begins”

obi wan: “let them pass between us”

anonymous doomed fighter pilot: “they’re all over me”

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.