NEW RELEASE:
Today we're releasing CortexMAE: a family of fMRI foundation models trained on 2.1K hours of open fMRI data.
We're also releasing Brainmarks: an open benchmark suite for evaluating fMRI foundation models.
Full paper is on arXiv (accepted to ICML 2026)
A thread:
‼️Οι μοναδικές στρατιωτικές φυλακές για παιδια στον κόσμο βρίσκονται στο Ισραήλ.
Τα ξυλοκοπούν, τα βασανίζουν και τα βιάζουν συνήθως!
Το ποσοστό καταδίκης των δικαστηρίων είναι 99,7%
Η μοναδική δημοκρατία στην Μέση Ανατολή........
“They want me to remain silent, that is what the world wants, right? Be silent, and don't say anything. I think it's completely the opposite. Why can’t I express what I feel? Just because I am a manager? Innocent people are being killed, I condemn all of them.”
—Pep Guardiola
A foreign ethno-nationalist movement from India is moving aggressively into U.S. politics. American discourse fixates on Chinese Communist Party interference and Russian disinformation. Hindutva — the supremacist Hindu nationalist movement that powers India's ruling government — barely registers in the conversation. Three of its collaborators sit in U.S. Congress.
Capitol Hill's Hindutva Squad — Suhas Subramanyam, Shri Thanedar, and Raja Krishnamoorthi — have spent years making common cause with the American arms of this movement. VHP America and CoHNA pushed H.Res. 69, the so-called "Hinduphobia" resolution. Thanedar authored it. Subramanyam and Krishnamoorthi cosponsored. The same network has funded all three of their campaigns and worked to kill state-level legislation built to protect Americans from transnational repression. Krishnamoorthi shared a stage with RSS chief Mohan Bhagwat at the 2018 World Hindu Congress in Chicago, then keynoted the RSS's birthday celebration in 2019 before a photograph of M.S. Golwalkar — the second RSS chief, whose 1939 book held up Nazi Germany's treatment of Jews as "a good lesson" for India. Later that year, Krishnamoorthi was the only Indian American member of Congress to attend Howdy Modi in Houston — a rally for a man then banned from the U.S. for over a decade for orchestrating the 2002 anti-Muslim Gujarat pogrom. Thanedar escorted Modi to his June 2023 joint address to Congress.
Then in February 2026, the middleman in India's plot to kill an American citizen on American soil pled guilty in a New York court. One month later, the U.S. Commission on International Religious Freedom recommended Washington sanction India's RSS — the paramilitary force at the heart of that government — for egregious violations of religious freedom. These three keep cozying up anyway. The full record is in the video.
@SuhasforVA, @ShriThanedar, @CongressmanRaja
#HindutvaSquad #Hindutva #HRes69 #VHPA #CoHNA #RSS #TransnationalRepression #USCIRF #ForeignInfluence #CapitolHill
@SamanthaTaghoy Deliberately leaving out persecution in Israel and India as it doesn't fit your islamophobic far right agenda... You don't care for no christians, you just wanna push an agenda...
Rayo Vallecano supporters group Bukaneros watched the basketball match between Iraklis BC and PAOK BC from the stands together with their friendly group from Iraklis, Gate 10.
In the stands, an Antifa flag is also seen together with the banners of the two groups. There is a friendship between the two groups going back nearly 20 years, and the basis of this bond is that both groups have an antifascist identity.
Gate 10 was also with Bukaneros a few days ago at the match between AEK Athens and Rayo Vallecano, and their banners were in the stands.
Someone just built a doomsday computer that runs without the internet.
It's called Project N.O.M.A.D. It packs a local AI, all of Wikipedia, offline maps, medical guides, and full Khan Academy courses into a solar-powered mini PC.
Runs on 15 watts. 100% Open Source.
I hear this every week in my office:
"Doc, my heart rate hits 150 during squats — that's cardio, right?"
No. And if your cardiologist hasn't explained why, keep reading. 🧵
Sometimes I feel outnumbered on X because most Greeks on here are pro-Israeli.
Then I get daily reminders that back in the homeland it’s mostly “free Palestine” 🇵🇸 💪🏼
@RagnLod@iraklishoops Το ότι δε θα έπρεπε εξαρχής να ακούγεται ο εθνικός ύμνος σε ματς συλλόγων σε ποιους το λέμε;
Να το πούμε στην εθνικόφρονα ΕΟΚ που έχρισε τον Γουοκαπ Έλλην; 🤣
@iraklishoops Η ομάδα έμεινε ξεκρεμαστη όταν δεν αντικατέστησε τον Φόρμαν. Παίζανε όλοι πάνω από ότι μπορούσανε για να καλύψουν κενά, με αποτέλεσμα κούραση, ντεφορμαρισμα και ήττες που υπό άλλες συνθήκες θα μπορούσαν να ήταν νίκες και ψυχολογία. Ψυχολογία που έλειπε γι'αυτό ήταν όλοι χαμένοι
@iraklishoops Η ομάδα δεν έπρεπε εξαρχής να είναι στο κυνήγι των 7 πόντων από την Πολωνία με πίεση. Μετά τον Παναθηναϊκό χαλάρωσα με πάρα πολύ, έχοντας ως δεδομένα τα πλέι-οφ και το Φ4, και από ότι φαίνεται θα μείνουμε εκτός και στα 2. Δεν πειράζει, η χρόνια ήταν πετυχημένη όπως και να χει
🚨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?