roni ilan retweeted
ווואי וואי וואי וואי.
הערבים נוהרים ולא רק הם.
יש תוצאות של קלפיות
הפער בין המעוזים גדל.
בבוקר מעוזי המרכז־שמאל הקדימו את מעוזי הליכוד ב־1.4 נקודות. בצהריים הפער כבר 4.2.
מעוזי המרכז־שמאל כבר השיגו 48.9% מאחוז ההצבעה הסופי שלהם ב־2022, מעוזי הליכוד רק 44.7%.
ובחברה הערבית זינוק: 48.5%!!!
שורה תחתונה למי שמסתבך עם גרפים ונתונים:
נכון לכרגע אחוזי ההצבעה מטורפים במיוחד אצל מצביעי אופוזיציה.
או במילים אחרות... זה נראה שיש טרוף לשינוי של הממשלה הנוכחית.
טפי טפו טפנ חמסה שום בצל
roni ilan retweeted
הפרויקט מתוכנן לקום במתחם שהיא אהבה במיוחד, בתכנון משותף עם אנשים על הרצף, במטרה ליצור מרחב חינוכי וחווייתי שינגיש לקהל הרחב את עולמם הפנימי, החושי והרגשי של ילדים ובוגרים עם אוטיזם.
כי אין דרך מדויקת יותר להמשיך את האור של נויה מאשר לקדם הבנה, סובלנות, אמפתיה וקבלה עמוקה של החוויה האוטיסטית כחלק מהמגוון האנושי.
ואין יותר ישראלי מלהתגייס ולתמוך בפרויקט הזה.
הקישור לתמיכה ותרומה ממש כאן:
jgive.com/new/he/ils/donatio…
הלוואי שנצליח להניע אפילו בקצת את גיוס התרומות ❤️
נויה וכרמלה - לא נשכח. יהי זכרן ברוך.
roni ilan retweeted
אף אחד בממשלה לא ייקח אחריות על הקטסטרופה המטורפת הזו. אף אחד. תזכרו מה אני אומר לכם. בכירים בצה״ל יתפטרו, חלקם יודחו, אבל לא יהיה אדם אחד בממשלה - לא ראש הממשלה, לא שר הביטחון, לא חברי הקבינט הביטחוני, שייקחו אלפית אחריות. כשהמלחמה הזו תיגמר ונספור את מתינו, תחל הבריחה מאחריות.
Just a quick reminder that the frontal lobe isn't fully developed until the early 30s. The level of impulsiveness is staggering
roni ilan retweeted
If you are a lab researcher, I urge you to consider what the next few years will actually feel like. Do you want to kick off a superintelligent RL run without a rigorous understanding of its mind? Should you put your head down because “it’s happening anyway” - or take this moment to call for different conditions?
roni ilan retweeted
If you're a female…
…with high cholesterol
…in your 30s
…you may be misdiagnosed
In one study, ~6% of women were labeled with high cholesterol… when they were tested again later in their cycle, their results came back ‘normal’.
Your cholesterol can swing by 19% depending on the day you get your blood drawn.
+ First half of cycle = higher.
+ Second half of cycle = lower.
Your cholesterol isn't one number. It moves monthly with your hormones. In the first half of your cycle, your cholesterol reads high. After ovulation, it falls, hitting its lowest right before your period starts.
When I found this out, I was shocked.
Why does no one know this?
The answer took me down a rabbit hole.
When your blood lab results come back, there's a "normal range" printed next to each number. That range is created by measuring a big group of people, and calling the middle 'normal.'
But for decades of blood draws, cycle days weren’t recorded… so their cholesterol variation was invisible.
Then, it got worse in 1977 when the FDA excluded women of childbearing age from early drug trials for 16 years.
So it wasn’t until 2010 when the first rigorous, large-scale proof came back that cholesterol changes with the cycle. A female health study looked at blood biomarkers and compared them with cycle data.
It took 60 years to figure that out.
And yet, it's 2026, another 16 years later, and the reference ranges are still outdated.
So… why hasn't this been fixed?
Heres 7 reasons why...
#1.
Most doctors don't know. It is not in their training or guidelines.
#2.
The lab companies haven't built for it. There's nothing stopping Quest or LabCorp from adding this. They set their own systems… they just haven't prioritized it. The individual who takes your blood has nowhere to note what cycle day you’re on.
#3.
Self-reported Cycle days. Most women were never taught to track their cycles. The education is thin, and full of myths (most women don’t ovulate on day 14). Even a woman who tracks carefully may have imperfect information. For example, ovulation shifts month to month, and sometimes it’s skipped.
#4.
Cycle phase confirmation. Cycle phase can be confirmed with a blood or urine hormone measurement alongside the draw. It's just that nobody has added this step.
#5.
The starter data already exists. The 2010 study measured hormones and cholesterol across hundreds of women's cycles. That's enough to build a first version of phase-specific ranges. Validating it at scale would take more data, but collecting it is now trivially cheap.
#6.
We repeat the test, but still ignore the cycle. Cholesterol is often checked more than once, which can help smooth out biological variation. But those repeat tests usually aren't timed to the menstrual cycle.
#7.
Ranges only change when someone decides to change them. There's no central body that sets reference ranges… each lab company adopts and verifies its own. When pediatric ranges turned out to be wrong, a research group rebuilt them from thousands of healthy kids and labs adopted the results. The same thing could be done here. It’s just that no one has started.
The data to fix this is completely collectible.
A third of American women already track their cycles on their phones.…. and the biobanks have millions of cholesterol results. Nobody has put the two halves together.
Insanely… cholesterol is not the only marker that varies across your cycle. In one study, about 50% of the molecules measured in women’s blood and urine moved with the cycle. And yet… it's still almost entirely unknown.
So I am running my own experiment.
I'm collecting 14 million data points on my menstrual cycle… my brain, blood, skin, saliva, grip strength, etc. We’re spending $2.6 million on this. And every second of my day is monitored and accounted for.
This will be the most quantified menstrual cycle ever.
We’ll post all results here (good and bad, wins and failures) with you for free.... sharing all the practical insights.
My hope is that we can put words and numbers to the female experience.
roni ilan retweeted
Yes and no. For many Physics problems where the fundamentals are well established and they can be reduced to applied Math, the answer is yes. Many others - including almost all of the most fundamental problems - require an entirely different paradigm before they can become amenable to mathematical problem solving. This is one of the deepest insights from the history of Physics that we have almost completely forgotten over the past half a century +. At some point AI may be able to provide us with such a deep insight, but I have not seen even a glimmer of such capabilities from the current models.
roni ilan retweeted
I’m getting tired of “experts” like this misunderstanding what they’re looking at.
LLMs are giant databases of stuff HUMAN BEINGS have done.
They are the EXHAUST of humanity.
Prompts are database queries into EXISTING DATA.
It’s a fuzzy search engine, not intelligence.
I am getting tired of reading 'experts' like LeCun repeatedly claiming that our AIs are nowhere near human-level intelligence.
Let us look at the evidence.
US universities rank students based on standardized tests like the SAT. Current AIs achieve near-perfect SAT scores. They also beat tests like the GRE.
A few years ago, it was notable when early ChatGPT scored ~120 on an IQ test, a common measure of human intelligence. An IQ of 120 is well above average. Current AIs reportedly have IQ scores similar to those of leading scientists.
It is not just in tests. I can ask an AI to produce a science paper that looks undistinguishable from what a PhD level student could do. I just have to give it the data. Better yet, from a prompt, agents can run the experiments and collect the data, and then write the papers.
Those of us who try to get work done with AI know what is possible. You can't possibly just say 'this is nowhere near human-level intelligence'.
In software, good AIs show a greater mastery of, say, C++, than your average software engineering professor. You could just build a formal test to prove it. The difficulty is that the professors would refuse to take your tests.
At this point point, someone will object 'yeah, but your AI can't do this simple thing that we can all do'. Fine. These AIs do not have *human* intelligence. They are very much not human beings. They are something like alien intelligence. They can code straight in assembly language, but have trouble counting characters in words. But that's the result of trade-offs. A dog or a monkey can solve some problems faster than you can.
But let us be fair. As a species, these AIs have definitively 'human-level intelligence'. You can't spend decades setting up cognitive tests for human beings, have these AIs beat us in these tests and then say 'well, that's not real intelligence'. Come on !
roni ilan retweeted
Sadly, this is happening everywhere:
"LLM fallacy: a cognitive attribution error in which individuals misinterpret LLM-assisted outputs as evidence of their own independent competence, producing a systematic divergence between perceived and actual capability."
roni ilan retweeted
That's us! 🌍
The Artemis II crew captured beautiful, high-resolution images of our home planet during their journey to the Moon. As @Astro_Christina put it: "You guys look great."

ALT A full disk image of Earth, as seen from the Orion capsule. The planet is a pale blue, swirling with white clouds and glowing slightly lighter blue in place from reflected light. From about 8 to 9 o'clock, a large brown landmass is Africa, with the Iberian peninsula twinkling with lights just where the planet curves. At the 1 o'clock spot, aurora glow in a thin green glow, just barely separated from the planet's surface. Earth is set against the black of space. Credit: NASA/Reid Wiseman
roni ilan retweeted
⛵️⛵️⛵️⛵️⛵️⛵️⛵️
אם יש לכםן מישהו שתקוע בחו״ל או רוצה לצאת מהארץ - אנחנו מתארגנים.
אני בעלת יאכטת מפרש וכמובן בכל הקבוצות של השייטים.
קחו בחשבון:
הגעה / חזרה מקפריסין - יום וחצי הפלגה.
זה יהיה בתשלום, ותלוי מזג אוויר.
דיאם פתוח. רטווטו.
🌊🌊🌊🌊🌊🌊🌊🌊
roni ilan retweeted
This is amazing - a periodic table showing every element and how it’s used.
Chemistry really is everywhere. 🔬✨
What’s your favorite element?
roni ilan retweeted
Scaling laws in AI – where do they come from?
The discovery of neural scaling laws several years ago showed that the loss decreases predictably as a power law in model size, amount of data, and compute. But why? And what sets the exponents of the power law?
The most popular explanation is that the dataset already has power law correlations in it (for example, power laws are prevalent in natural language corpora, e.g. Zipf’s law, etc), which translate to power laws in the loss.
We studied transformers performing next token prediction on sequences coming from random walks on random graphs, where the data has no power law correlations. Nevertheless, after training the model, we observed power laws in the loss that look similar to those found in natural language. For example, here are results from a random walk on an Erdös-Renyi graph with 8K edges and 50K nodes:
This challenges existing explanations, since this dataset of random walks falls outside of the assumptions made in existing models of scaling laws. Going forward, we need explanations of scaling laws based on expressivity and learnability of discrete data, where there is no data manifold, and which do not require the data to already have power laws built in.
We also found a setting where we could tune the complexity of a language dataset by starting with a bigram model and gradually dialing up complexity until we get to natural language. This allowed us to track how the exponents of the scaling laws change with complexity:
roni ilan retweeted
בקשה אישית מאביו של מילאר ז"ל שנרצח במג'דל שמס: מילאר חלם על שלום עולמי. הוא מבקש להפיץ את הסרטון ואיתו את התקווה להגשמת משאלתו של הבן שנרצח עם עוד 11 ילדים וילדות.
roni ilan retweeted
Heartbroken to share that Yagev Buchshtab, brother of PhD student Nufar, was murdered in Hamas captivity. Too much unbearable pain. This should not have ended, or started, this way.
roni ilan retweeted
בוקר שמחת תורה. ב6:29 רצנו לממד והייתי בלי הטלפון. 18 דקות אחר כך הייתה הפסקה רגעית במטחים ויכלתי לרוץ להביא את הטלפון ולצייץ.
teapawt.org/smolan_with_s/status/1…



































