eyes full of stars

sam | 29 | she/they spotify baddestbitchcas

[sivavakkiyar]

posting is like bloodletting

[stephsxcorner]

I have this very strong opinion that when you're young you feel things on such a intense and detailed level...

[lakevida]

i would be so bad at performing surgery thank god i dont have to do that

[ostolero]

[kosmogrl]

[enki2]

[overwelmd]

cocorrina

[missjamiekaye]

Kitties in the Garden. A little gift I made for a friend of her little visitors.

[klfg]

[source]

[klfg]

🚨BREAKING: OpenAI published a paper proving that ChatGPT will always make things up.

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.

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

[kat4884]

the gooners

they’re everywhere

[felixcloud6288]

Dat's right boss. Now just tell us what ya need and we'll get it done boss.

[quadtator-lawful-evil]

who needs ta sleep with da fishes boss

[bananapudding752]

sorry bawss, I was too busy searchin up da anime girls boss. real pretty.

[felixcloud6288]

Forget dem anime goils. Dere only temporary. But da boss is forever. Right boss?

[hellsitegenetics]

String identified: t g t’ at' gt . t t at a a ' gt t . ta t a a, a t ac a a g . a tt. gt a g. ta. t a . gt ?

Closest match: Oeneis jutta genome assembly, chromosome: 18 Common name: Jutta Arctic

(image source)

[felixcloud6288]

Da boss got turned into a butterfly!!

[pfaugh]

Wadda bea-utiful transformation, boss. goes to show youse gotta preciate the lil things in life

[hraeiou]

ay boss congrats on da metamoiphosis

[wrathofachilleus]

born to be an abstract concept, forced to be a percievable entity

[walk-up-songs]

[one single bloodcurdling agonized scream] ok time to lock in

[dyrdeer]

for everyone asking, tumblr is making it impossible for you to add onto ANY post without making it its own separate post. If you put any addition on someone's post its now your post. The OP won't even be privy to any notes it gets unless they check that specific post. This makes it impossible to track anything which is really detrimental to artists who need to know their reach or just see what people are saying about their work. It also just means you can't reblog with any additions without taking notes away from your original post. This is part of what made twitter hostile and its the last gasping breath of a site run by people who don't understand it.

[starvy]

never stop bisexualing

[daenerys-targaryen]

love blogging and blogging and being on my blog and updating my blog and blogging at the same time as other bloggers and blogging. I heart My Blog. blog also sounds like bog which is fun. Love bogging with you all.

[daenerys-targaryen]

tumblr staff:

[junkratsadstuck]

here at the sandwich shop, we’ve started to notice some people who are new to sandwiches aren’t used to meat and cheese between two slices of bread. they find this practice strange and confusing. that’s why we’ve decided to cut the bread out all together. from now on, we’ll just serve slices of meat and cheese on a plate.

[junkratsadstuck]

we know that many of our loyal sandwich shop customers have been coming here for years to buy our delicious sandwiches. but some people don’t “get” sandwiches, and we need to try and appeal to them with an easier-to-understand meal format. we will no longer be serving sandwiches. all of our food will just be cold cuts on a paper plate. we love our customers and appreciate your understanding <3

[derinthescarletpescatarian]

Yes we know Craig's Cold Cuts down the street does cold cuts better than we do and has millions of customers. That's why we need to switch to cold cuts, to get their customers. We hope that our long-time sandwich loving customer base will be patient in this transition to cold cuts and welcome the certain influx of Craig's Cold Cut customers who will surely come here to enjoy our worse cold cuts with you, our sandwich-deprived customer base, whom we value so much.

[derinthescarletpescatarian]

I love how this goes around again every time the tumblr staff does something fucking stupid.

[potatomode]

They soulmated too hard that their stylists accidentally synced up