manv

star-stuff making stuff

[manv @vmanvs]

I was adrift and called it being lost. I was held by the deep and called it abandoned. I waited for land that never came and when I stopped waiting, I finally felt the water holding me, and knew: I had been home the whole time, raging against the gift of it.

[manv @vmanvs]

@JordanNanos I am more interested as to how much this touched nvidia ip while they were integrating rtx spark

[manv @vmanvs]

@spacexai how cool would it be if you also give gpus for us to tinker on, just like how @GoogleColab promised to give us gpus on ai pro plans (i haven't received them yet btw)

[manv @vmanvs]

@jino_rohit everyone brother, everyone be tryin

[manv @vmanvs]

@alexandr_wang @dtrinh banger

[manv @vmanvs]

@nutlope thank you

https://t.co/tW6xrLQLL2

[manv @vmanvs]

@nutlope someone do this for X bookmarks too

[manv @vmanvs]

@finkd sparks of a comback

[manv @vmanvs]

@prathamgrv @TensorTonic whooaa, i could see the pile of cash from here

[manv @vmanvs]

@yacineMTB do what torvalds did with git

[manv @vmanvs]

@nutlope someone do this for X bookmarks too

[manv @vmanvs]

@tenderizzation banger

[tender @tenderizzation]

didn’t take the contrarian allegations seriously and then discovered today that my phone autocompletes “very nice” with “but”

[manv @vmanvs]

@ParamSiddh dinosaurs

[manv @vmanvs]

@neural_avb looks interesting

[manv @vmanvs]

@yacinelearning would you be sharing that alpha? kindly do so 🥺

[manv @vmanvs]

@oprydai when you go after dji, dji simultaneously comes after you

[manv @vmanvs]

turns out attention indeed was all you needed

[Superman @thesupermannx]

Researchers mathematically proved Transformers are computationally universal.

They applied the concept of Kolmogorov complexity, the mathematical absolute limit for perfect data compression and algorithmic generalization.

And it proved that Transformer encoders are computationally universal.

In plain English: they proved that there is a mathematical state where an AI model becomes perfectly compressed and generalizes flawlessly to new data.

The architecture is theoretically flawless.

But there is a catch. And it changes how we look at every AI lab on earth.

When the researchers built a specific objective to force the AI to find this mathematically optimal, perfectly generalized state... the Transformer was entirely capable of holding it.

But the optimizer couldn't find it.

Standard training methods, the exact same optimizers used to build ChatGPT, Claude, and Gemini, completely failed to discover this perfect solution from a random start.

We are staring at a massive, structural disconnect.

Transformer architecture is mathematically capable of universal, optimal intelligence. The engine is perfect.

But the steering wheel is broken.

Our training methods are simply too primitive to unlock what the model is actually capable of.

We don’t need to invent a new architecture to reach the next leap in AI reasoning.

Intelligence is already buried inside the models we have right now.

[manv @vmanvs]

@neogoose_btw mathematicians before "internal model"

[manv @vmanvs]

@barrowjoseph flops are readily available, bandwidth is not

[manv @vmanvs]

@0xSero

[manv @vmanvs]

@maharshii been there... and always will be