0-_-0

In 5 years phones can easily have 1 TB flash

You don't think you could you the same to a human?

On what?

Depends on the road:

A 9th-power law is most predictive when the road is barely able to withstand the 6T load; and the per-crossing damage is roughly linear to axle-weight when the pavement is able to withstand much higher loads than 6T per axle.

From TFA:

“I’ve entered prompts into AI before,” he says. “This is the first time AI entered a prompt into me.”

How about chess?

What if there are math problems that can only be brute forced? How do we know that Navier-Stokes even has a nice and elegant solution?

Is there an example of foveated rendering implementation?

Instruct the AI to output an .MD file that can be used to reconstruct the code

Why didn't Russia just use local models?

You're hallucinating

Affected by what exactly?

The kv cache is not discretised, only the output tokens.

Reasoning is already taking place in latent space, the CoT is just a sideeffect

I wonder if you could prune the kv cache as well

Same is true for 1

Right, so it was sealed only in the biological sense, which makes sense since there needs to be an inflow of energy

No but they could over 5 million years

The information that was missing from my life:

For the Average Man, gaining an inch of height is worth losing roughly $25k in annual salary. Meanwhile, an inch of pecker is roughly offset by reducing annual salary by $128k.

You would need to hide the existence of the original profile while in the decoy profile for this to work, which GrapheneOS considers too complex to implement

It's not a benchmark, it's a meme

They want to be power efficient, while Nvidia doesn't. That means a much higher cost (larger chip area) for the same performance.

Yes, but all subsequent tokens do depend on them, so they are still being charged for

Don't you jump the highest with com_maxfps 333 anyway? Or did Quake Live change this from Quake 3?

You can only see satellites during twilight when they can reflect sunlight. Don't panic.

In that article Grok shows highest neutrality

This made me wonder about a 3D printer alternative that builds things by folding a thin sheet of metal into arbitrary shapes instead of extruding filament.

Even if you had all that you would get completely different weights out at the end, and you also don't have the resources to "compile" an LLM because the compilation can cost $100M. If you were given the training data but not the weights, would you consider that open source?

Source code makes it easier to modify a binary, training data doesn't make it easier to modify a trained network.

If we can launch 1M satellites, how many telescopes can we launch?