Sabitlenmiş Tweet
Mary
2.5K posts

Mary retweetledi

I know a lot of people in the UFO community truly believed in President Trump, I wonder how those people feel right now after he publicly trolled the entire concept of disclosure with a fake build up of theatrics only to reveal Aliens.gov is about illegal immigrants
English
Mary retweetledi
Mary retweetledi
Mary retweetledi

I literally haven’t typed anything in weeks
Since I started using Grok’s speech-to-text in Hermes Agent...
I just talk
It transcribes everything perfectly
Every word. Every single time
My thoughts flow straight from my mouth to Grok...
And it gets it right. Every time
Once you experience this...
You won’t go back
English

@TheProjectUnity The pope is going to feel sooo bad when he wakes up to god.
English

Opinion: Unlike the US president, the pontiff is choosing to grapple with the serious challenges of AI. ft.trib.al/V1Gyr6w

English
Mary retweetledi

Sketchy facts about Peter Thiel:
-His name is literally an anagram for "The Reptile."
-Is ostensibly a libertarian but built infrastructure for the modern day surveillance state
-Has an unhealthy fascination with the "Antichrist."
-named his company Palantir after the seeing stones in Lord of the Rings corrupted by the dark Lord Sauron
-genuinely looks like a supervillain

Latinometrics@LatamData
ARGENTINA | Tech billionaire Peter Thiel relocates to Argentina, citing concerns about the U.S. future and shared ideology with leadership. (NYT)
English
Mary retweetledi

Quantum entanglement is one of the most experimentally verified and least intuitively understood phenomena in all of science. Einstein hated it. He called it "spooky action at a distance" and spent years trying to prove it could not be real. The experiments proved him wrong.
Here is what actually happens. Two particles interact and become entangled, meaning their quantum states are linked. Separate them by any distance, measure one, and the other instantly reflects a correlated result, regardless of how far apart they are. Not after a delay. Not at the speed of light. Instantly.
This does not violate relativity, and that distinction matters. No information is actually transmitted between the particles. You cannot use entanglement to send a signal faster than light because the outcome of each measurement is still random. The correlation only becomes apparent when you compare results through normal, slower-than-light communication. The universe found a loophole and closed it at the same time.
What has been confirmed experimentally, most decisively through tests of something called Bell's theorem, is that the particles cannot have simply carried hidden predetermined values with them from the start. The connection is real and it is nonlocal. The particles genuinely do not have definite states until one of them is measured, and once one is, both are.
Entanglement has now been demonstrated across distances of over 1,200 kilometres using satellite-based experiments. It is the backbone of quantum computing and quantum cryptography, both of which are being actively developed and deployed right now.
The physics is settled. The explanation for why the universe behaves this way is not.
That gap between what we can prove and what we can understand is where physics gets genuinely uncomfortable.

English
Mary retweetledi

JUST IN: Elon Musk explains why SpaceX built its own AI training software from scratch.
Most AI companies use Google's tools to train their models. The problem is those tools were built for Google's chips, not Nvidia's.
SpaceX runs 220,000 Nvidia chips. Using the wrong software is like putting the wrong fuel in a race car.
So they wrote their own. In C, one of the fastest programming languages that exists.
Elon Musk says they might even go one level deeper and write parts of it in assembly, the closest thing to speaking directly to the hardware.
"JAX is mostly optimized for small to medium-sized training runs on Google TPUs, whereas we need massive training runs on Nvidia GPUs."
The result is training speeds over 10x faster than what the industry standard delivers. Faster training means smarter AI, built at a fraction of the cost and time.


Elon Musk@elonmusk
Yes. It’s not that we’ve discovered some magic bullet, but rather that JAX, or at least the open source version of it, is mostly optimized for small to medium-sized training runs on Google TPUs, whereas we need to massive training runs on Nvidia GPUs. Pipeline parallelism is essential and crushes fully-sharded data parallelism at scale. And C will compile to the most efficient binary short of assembly. Maybe we will do a little assembly too.
English
















