@fromdevoid @Jonathan_Blow Ah you should probably not watch then. Please warn all your friends not to watch, too.
English
C. Aaron Rodgers
519 posts
C. Aaron Rodgers
@CAaronRodgerus
The numbers lead a dance. - Sting
เข้าร่วม Nisan 2023
1K กำลังติดตาม123 ผู้ติดตาม
Not entirely sure I want to know what a computer programmer has to say about the simulation hypothesis, even when it seems to align with my own conclusions on the matter.
Unless... are you going to base your argumentation on existing debates by physicists, philosophers, cosmologists and so forth (certainly with your own views added)? Or is this going to be merely an argument of faith, like the whackjobs swearing by it?
English
As someone who programs game ("simulation") engines, I have a long-term project where I want to explain why the "simulation hypothesis" is nonsense in like 5 different ways (while pointing toward what a sensical version would look like).
So this Saturday at 10am Mountain Time, I will do a livestream that is the first in the series of Simulation Saturdays. To start with, I just want to lay out the framework in a not-organized, random-discussion kind of way, to just make an outline of what all the relevant topics would be. Then on later Saturdays we can go to various subtopics and talk about them.
(Unlike #screenshotsaturday, the first Simulation Saturday will be at a time that is widely recognized as Saturday by many people, though I make no promises about subsequent Simulation Saturdays.)
English
@Jonathan_Blow 👀 Are physics simulation demos (rigid body, fluid, orbital mechanics) under consideration?
English
I honestly believe that Vim Motions don't actually make you faster at using a text editor, and a well implemented multiple-cursor system + typical modifier keys are actually much faster in practice.
Evidence: watch me program on streams with Sublime Text and judge for yourself.
English
BREAKING: Just five minutes before Trump's announcement to halt the attacks on Iran, massive trades reportedly hit the market.
In one move, $1.5 billion in S&P 500 (ES) futures was bought while $192 million in oil (CL) futures was sold.
These orders were 4–6x larger than anything else at the time.
The trader seemingly made huge gains.
Unusual.
English
@KSHartnett @leanprover @mathematics_inc @dwarkesh_sp Is it really the community? I've only seen angry remarks by one individual.
English
Reading this morning about the at times angry response of the @leanprover Mathlib community to @mathematics_inc's sphere packing formalization, I'm reminded of something Tao said on @dwarkesh_sp recently.
"I think in the future, there will be entire professions of mathematicians who might take a giant Lean-generated proof and do some ablation on it, trying to remove parts of it and find more elegant ways. They might get other AIs to do some reinforcement learning to make the proof more elegant, and maybe other AIs will grade whether this proof looks better or not."
Dwarkesh Patel@dwarkesh_sp
Terence Tao explains the beauty of Lean proofs. Even if they’re not very comprehensible on their own to humans, they can be analyzed more easily - each bit of the proof can be taken apart, analyzed, tweaked, and understood in terms of how it fits into the whole.
English
@AbdelStark How have you chosen to differentiate between deterministic operations that are most efficiently run on the GPU via mat-muls, and which ones run most efficiently as compiled code on the CPU?
English
Can LLMs be PROVABLE computers?
Percepta showed that a transformer can BE a computer. Compiled weights, deterministic execution, 30k tokens/sec.
But nobody asked the obvious follow-up: how do you know it computed correctly?
So I built the verification layer. A STARK that proves it 👇
Christos Tzamos@ChristosTzamos
1/4 LLMs solve research grade math problems but struggle with basic calculations. We bridge this gap by turning them to computers. We built a computer INSIDE a transformer that can run programs for millions of steps in seconds solving even the hardest Sudokus with 100% accuracy
English
@littmath in math you can immediately verify that a supplied proof is correct. so a million wrong guesses don't matter. it's the opposite for arrangement of physical atoms. specifying correct is near impossible, and wrong guesses are costly.
English
Given what current-gen LLMs (say, in math, but whatever) can do, I think their apparent limitations are kind of mysterious. What is the blocker preventing, at present, high quality fully autonomous work?
English
@colin_k Will check for linux compat when I get home! This lights up all the right neurons for me.
English
🚁Cleared Hot is a daily deal on Steam today !! We're featured on the front page of the store 🤯
We will be raising the price before the next major update. If you've been waiting, now's the best time to grab it.
The remaining chapters (2 and 3) will be free updates.
English
@andrew_n_carr Can't parse it yet. Is the primary application graphics? CFD?
English
This is the coolest thing I have seen in an exceedingly long time.
Simone Foti@simo_foti
It's time to bring 3D meshes into modern machine learning properly! 🛸 Our work solves the non-differentiability of the Exp map on meshes, enabling gradients to flow directly through geodesics. It’s differentiable, GPU-fast, and fully parallelised. circle-group.github.io/research/DSG
English
@Techjunkie_Aman Didn't you hear, nobody wants systemd anymore. Ill be switching away at the earliest convenience until such a day this is no longer the case. @CachyOS lets goo!
English
CachyOS is basically Arch Linux on steroids.
Instead of using normal builds, it’s made specifically for your CPU.
So everything runs smoother.
• x86-64-v3 / v4 optimized packages
• Zen4 instruction tuning
• LTO + PGO + BOLT optimizations
• custom kernel with tuned EEVDF scheduler
• optional schedulers (BORE, BMQ, RT, sched-ext)
• 17+ desktop environments (KDE, GNOME, Hyprland, i3, XFCE…)
It also has:
• custom performance kernel
• lots of desktop options (KDE, GNOME, etc.)
• easy installer (no complicated setup)
No tweaking.
No coding.
Just install and enjoy speed
English
@georgecurtiss Yeah we know. And that means the u64/u128 discrepancy in Duration will slap new learners in the face for all eternity.
English
@wavefnx Rust already going the way of C++ because they prioritize backward compatibility above all. See, for example, the u64/u128 discrepancy in Duration. Other languages that are more nimble and chase excellence over placating their user base will have more success.
English
Rust
All robotics firmware will be written with it and target STM32's, accompanying the CAN and other peripheral libraries I already have
If you need deterministic performance and uptime in a software application, you need x1000 more of it in robotics
Elon Musk@elonmusk
English
This is a 97 GB GeoTIFF in QGIS 4. No tiling, just dropped the image into the app and it loaded right away.
English
@joemccann This is great, and a welcome improvement, but why would we expect this to compare even remotely on performance with calling out to a native library?
English
This is actually insane.
Dude hard-coded a WebAssembly (WASM) interpreter into the weights of a transformer, losslessly.
In essence, a computer is running inside a LLM that can actually run computations, not infer or guess a calculation like most do today.
Christos Tzamos@ChristosTzamos
1/4 LLMs solve research grade math problems but struggle with basic calculations. We bridge this gap by turning them to computers. We built a computer INSIDE a transformer that can run programs for millions of steps in seconds solving even the hardest Sudokus with 100% accuracy
English
@ChristosTzamos How do execution time and power consumption compare between this model and a native program written in, say, C?
English
4/4 Read more at our blog post:
percepta.ai/blog/can-llms-…
English
1/4 LLMs solve research grade math problems but struggle with basic calculations. We bridge this gap by turning them to computers.
We built a computer INSIDE a transformer that can run programs for millions of steps in seconds solving even the hardest Sudokus with 100% accuracy
English
Another blog post in the HoTT series.
bartoszmilewski.com/2026/03/10/the…
English
hey its been 2 months
guess we dont need humans at all anymore!
Tom Blomfield@t_blom
This will remain true for about 2 more months.
English
