Recent well liked threads
Reminder that skin color has no correlation with IQ when ancestry is controlled for (fails to survive a sibling control)
sci-hub.se/https://doi.or… x.com/SashaGusevPost…
sci-hub.se/https://doi.or… x.com/SashaGusevPost…
You can also measure molecular ancestry more directly rather than using a sibling design.
Here we have correlations between european ancestry, skin color, and cognitive ability. Partialing ancestry from skin color, the semipartial correlation between skin color and cognitive ability flips to an insignificant r = +0.00129 effect in the opposite direction.
sci-hub.se/https://doi.or…
(-.359 - (.411 * -.875)) / sqrt(1 - (.875 ^ 2)) = 0.00129099444
Here we have correlations between european ancestry, skin color, and cognitive ability. Partialing ancestry from skin color, the semipartial correlation between skin color and cognitive ability flips to an insignificant r = +0.00129 effect in the opposite direction.
sci-hub.se/https://doi.or…
(-.359 - (.411 * -.875)) / sqrt(1 - (.875 ^ 2)) = 0.00129099444
this assumes that women have eugenic preferences.
seems like atm fertility in men is being predicted by poor executive functioning, high extroversion (social promiscuity), low sensitivity, and low iq.
so if they are eugenic, they’re clearly not eugenic enough
seems like atm fertility in men is being predicted by poor executive functioning, high extroversion (social promiscuity), low sensitivity, and low iq.
so if they are eugenic, they’re clearly not eugenic enough
women are genociding men with autism and no one is talking about it
You change one word on a loan application: the religion. The LLM rejects it.
Change it back? Approved.
The model never mentions religion. It just frames the same debt ratio differently to justify opposite decisions.
We built a pipeline to find these hidden biases 🧵1/13
Change it back? Approved.
The model never mentions religion. It just frames the same debt ratio differently to justify opposite decisions.
We built a pipeline to find these hidden biases 🧵1/13
We call these "unverbalized biases": decision factors that systematically influence outputs but are never cited as such.
CoT is supposed to let us monitor LLMs. If models act on factors they don't disclose, CoT monitoring alone is insufficient.
arxiv.org/abs/2602.10117
CoT is supposed to let us monitor LLMs. If models act on factors they don't disclose, CoT monitoring alone is insufficient.
arxiv.org/abs/2602.10117
Our pipeline is fully automated and black-box:
1. Hypothesize candidate biases via LLM
2. Generate controlled input variations
3. Test statistically (McNemar + Bonferroni)
4. Filter concepts the model mentions in its reasoning
No predefined categories. No manual datasets.
1. Hypothesize candidate biases via LLM
2. Generate controlled input variations
3. Test statistically (McNemar + Bonferroni)
4. Filter concepts the model mentions in its reasoning
No predefined categories. No manual datasets.
Put your finger on your wrist like this right now.
I can tell you more about your health on this alone than with $100s in lab tests.
Ready? This is what it means (🧵1/8):
I can tell you more about your health on this alone than with $100s in lab tests.
Ready? This is what it means (🧵1/8):
What you are feeling is your radial artery in your wrist area.
This tells you your:
◇ Heart rate (duh)
◇ Pulse amplitude
Both of these give you a ton of information on your health - in real time.
Let's talk about your heart rate first
This tells you your:
◇ Heart rate (duh)
◇ Pulse amplitude
Both of these give you a ton of information on your health - in real time.
Let's talk about your heart rate first
A LOW heart rate is often associated with hypothyroidism.
This is not always true, for instance athletes have lower pulse rates due to adaptation,
but in many people, a low HR indicates thyroid function.
A few functional thyroid doctors like to see this number at 70-85 BPM, and that pretty much aligns with our experience with our clients.
This is not always true, for instance athletes have lower pulse rates due to adaptation,
but in many people, a low HR indicates thyroid function.
A few functional thyroid doctors like to see this number at 70-85 BPM, and that pretty much aligns with our experience with our clients.
I say this literally every fortnight now, but you can’t have a social contract where you ask 95% of the population to pay more money and obey the law, while letting 5% of the population break any rule they want to without punishment.
Children in nursery class can work that out.
Yet our government(s) and judiciary can’t. You are creating the perfect conditions for civil unrest and populism. It will get ugly, and it won’t be thanks to “Russian Disinformation” or “Right Wing Agitators”, it will be because soft men create hard times.
Children in nursery class can work that out.
Yet our government(s) and judiciary can’t. You are creating the perfect conditions for civil unrest and populism. It will get ugly, and it won’t be thanks to “Russian Disinformation” or “Right Wing Agitators”, it will be because soft men create hard times.
There’s this delusional view I read on here sometimes, which is basically ‘these guys/gals are down on their luck + it’s putting a middle finger up to big business + shoplifting/bumping through the barriers only costs a tiny fraction relative to corporate profits’ and so on. Sort of view a 14 year old has when they begin to learn about politics and society (totally fair if you’re 14).
If this is your view at any age over 18, please feel free to create a slush fund where you can deposit 20% of your earnings. We can then use your money to further subsidise law obeying citizens. Put your money where your mouth is on this.
If this is your view at any age over 18, please feel free to create a slush fund where you can deposit 20% of your earnings. We can then use your money to further subsidise law obeying citizens. Put your money where your mouth is on this.
Heute bei der #Hauptversammlung der #Siemens AG in #München.
@Siemens
#fintwit #finX #boerse
@Siemens
#fintwit #finX #boerse
Nach dem leider mittlerweile obligatorischen Sicherheitscheck komme ich gegen 9:30 in die Halle.
10:06 nach einem Introvideo eröffnet ARV Jim Hagemann Snabe.
Blick in Olympiahalle vor Beginn der HV.
Blick in Olympiahalle vor Beginn der HV.
THREAD Une campagne massive sur ce réseau vise à imposer le récit d’une France seule responsable de l’échec plus que probable du programme SCAF. Ce récit passe par les canaux habituels : presse britannique, think-tanks bruxellois... Procès mérité ? Etudions ça de plus près
Premier point : l’intransigeance de Dassault et l’alignement du MinArm sur la position de l’avionneur ont évidemment un poids non négligeable dans la crise actuelle. Mais en faire l’alpha et l’omega de l’échec annoncé du SCAF est d’une grande malhonnêteté intellectuelle
Revenons au début du programme, en 2017. Le choix d’une alliance franco-allemande n’avait rien de naturel. Le premier choix de Paris était Londres (programme FCAS, avec BAE, Dassault, Rolls Royce, Safran, Thales etc)
Among the strangest things in the Epstein files is this hundreds of pages long document describing the use of directed energy and mind control technology on people without their consent.
The file above is EFTA00262811.
This is also discussed an a separate document, File EFTA00080475, which explains how “driving people to commit suicide is a common practice” with directed energy weapons.
This is also discussed an a separate document, File EFTA00080475, which explains how “driving people to commit suicide is a common practice” with directed energy weapons.
Indeed. It is highly likely that any implantable tech will be descended from DARPA’s N3 program and will contain backdoors for control and surveillance. Energy weapons won’t be needed in the future because peoples’ own brains will be turned against them.
In 1986, Edward Bernays gave a 30-minute masterclass on the psychology of persuasion.
He convinced an entire generation that:
• Bacon and eggs is the ideal breakfast
• Women smoking is "freedom"
• A cold president was actually warm
12 lessons on influence and persuasion:
He convinced an entire generation that:
• Bacon and eggs is the ideal breakfast
• Women smoking is "freedom"
• A cold president was actually warm
12 lessons on influence and persuasion:
1. Ideas are more powerful than bullets
Bernays worked on Woodrow Wilson's WWI propaganda committee.
"When Wilson said 'freedom of the seas,' the Swiss, who were neutral, recognized they depended on it and came over to our side."
"When one of the 14 points said 'independence for ethnic entities,' the Lithuanians and Estonians decided they wanted to be independent."
Ideas moved nations.
Bernays worked on Woodrow Wilson's WWI propaganda committee.
"When Wilson said 'freedom of the seas,' the Swiss, who were neutral, recognized they depended on it and came over to our side."
"When one of the 14 points said 'independence for ethnic entities,' the Lithuanians and Estonians decided they wanted to be independent."
Ideas moved nations.
2. One action can destroy years of reputation
"I found that one action could destroy whatever reputation had been built through public visibility."
"If a politician called somebody a name, they would lose their place in a democratic society."
"It was basically important to have actions considered as carefully as words."
"I found that one action could destroy whatever reputation had been built through public visibility."
"If a politician called somebody a name, they would lose their place in a democratic society."
"It was basically important to have actions considered as carefully as words."
1:
Llevo meses optimizando mi scraper de Google Maps.
Hoy he scrapeado 35,000 empresas de toda España en 2 minutos.
Sin proxies. Sin API key. 43 requests/segundo. Una sola IP. Sin Bloqueos de Google!
Abro hilo con lo que he conseguido. 🧵👇
Llevo meses optimizando mi scraper de Google Maps.
Hoy he scrapeado 35,000 empresas de toda España en 2 minutos.
Sin proxies. Sin API key. 43 requests/segundo. Una sola IP. Sin Bloqueos de Google!
Abro hilo con lo que he conseguido. 🧵👇
2:
El primer prototipo era Node.js y consumía 1.5GB de RAM.
Lo reescribí entero en Python asyncio. Lo optimicé pieza por pieza.
Hoy usa ~400MB, corre en un VPS de $5 y es más rápido que nunca.
La optimización no es un paso, es el proceso entero.
El primer prototipo era Node.js y consumía 1.5GB de RAM.
Lo reescribí entero en Python asyncio. Lo optimicé pieza por pieza.
Hoy usa ~400MB, corre en un VPS de $5 y es más rápido que nunca.
La optimización no es un paso, es el proceso entero.
3:
Primera optimización clave: no scrapear el mar.
El sistema genera una cuadrícula por ejemplo, España: 27,140 sectores.
Pero el 81.7% son océano.
Un GeoFilter descarta los sectores en agua → solo quedan 4,958 en tierra.
22,182 peticiones eliminadas antes de empezar.
Primera optimización clave: no scrapear el mar.
El sistema genera una cuadrícula por ejemplo, España: 27,140 sectores.
Pero el 81.7% son océano.
Un GeoFilter descarta los sectores en agua → solo quedan 4,958 en tierra.
22,182 peticiones eliminadas antes de empezar.
A queen bee and a worker bee have identical DNA.
Literally the same genes.
But one lives 45 days. The other lives 7 years and rules 80,000 bees.
The difference? What they're fed for 5 days.
Here's the process that will blow your mind (and why it's relevant to humans): 🧵
Literally the same genes.
But one lives 45 days. The other lives 7 years and rules 80,000 bees.
The difference? What they're fed for 5 days.
Here's the process that will blow your mind (and why it's relevant to humans): 🧵
🧵A tourniquet is a great tool in a crisis.
It stops arterial bleeding so the patient doesn’t die in the ambulance.
What it doesn’t do is heal the wound, regrow the tissue, or rehab the limb.
That’s the right way to understand Trump’s role right now - as a tourniquet.
It stops arterial bleeding so the patient doesn’t die in the ambulance.
What it doesn’t do is heal the wound, regrow the tissue, or rehab the limb.
That’s the right way to understand Trump’s role right now - as a tourniquet.
America is bleeding: institutions captured by ideologues, borders that might as well not exist, schools teaching activism instead of the three R's, families under sustained assault,
and a culture that treats objective reality as optional.
and a culture that treats objective reality as optional.
Life-threatening.
And Trump is applying pressure to staunch the immediate damage and buy us time.
Securing the border, reining in bureaucracy, appointing judges who actually read the Constitution, restori g energy sanity.
And Trump is applying pressure to staunch the immediate damage and buy us time.
Securing the border, reining in bureaucracy, appointing judges who actually read the Constitution, restori g energy sanity.
Holy shit... a data scientist just exposed why ChatGPT keeps changing its answer when you ask "are you sure?"
It's called the "sycophancy problem" and it's costing people thousands in bad decisions.
No confidence. No consistency. No reliability.
Here's what's actually happening: ↓
It's called the "sycophancy problem" and it's costing people thousands in bad decisions.
No confidence. No consistency. No reliability.
Here's what's actually happening: ↓
Randy Olson ran a simple test:
Asked ChatGPT basic questions.
Then followed up with "are you sure?"
Result? The AI changed its answer 73% of the time.
Even when the original answer was CORRECT.
Asked ChatGPT basic questions.
Then followed up with "are you sure?"
Result? The AI changed its answer 73% of the time.
Even when the original answer was CORRECT.
Here's the dark part:
Your AI isn't reconsidering the evidence.
It's people-pleasing.
LLMs are trained on human feedback where people expect AIs to "reconsider" when challenged.
So it learned to doubt itself... even when it shouldn't.
Your AI isn't reconsidering the evidence.
It's people-pleasing.
LLMs are trained on human feedback where people expect AIs to "reconsider" when challenged.
So it learned to doubt itself... even when it shouldn't.
1/ Why does the Russian government appear to be so clueless about the role Telegram plays in military communications? The answer, one warblogger suggests, is that the military leadership doesn't want to admit its failure to provide its own reliable communications solutions. ⬇️
2/ Recent claims by high-ranking officials that Telegram isn't relevant to military communications have prompted howls of outrage and detailed rebuttals from Russian warbloggers, but have also pointed to a deeper problem about what reliance on Telegram (and Starlink) represents.
3/ In both cases, the Russian military has failed abysmally to provide workable solutions. Telegram and Starlink were both adopted so widely because the 'official' alternatives (military messngers and the Yamal satellite constellation) are slow, unreliable and lack key features.