HackerTrans
TopNewTrendsCommentsPastAskShowJobs

0-_-0

4,338 karmajoined hace 8 años

comments

0-_-0
·hace 4 horas·discuss
They want to be power efficient, while Nvidia doesn't. That means a much higher cost (larger chip area) for the same performance.
0-_-0
·hace 9 horas·discuss
Yes, but all subsequent tokens do depend on them, so they are still being charged for
0-_-0
·anteayer·discuss
Don't you jump the highest with com_maxfps 333 anyway? Or did Quake Live change this from Quake 3?
0-_-0
·hace 3 días·discuss
You can only see satellites during twilight when they can reflect sunlight. Don't panic.
0-_-0
·hace 5 días·discuss
In that article Grok shows highest neutrality
0-_-0
·hace 7 días·discuss
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.
0-_-0
·hace 8 días·discuss
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?
0-_-0
·hace 8 días·discuss
Source code makes it easier to modify a binary, training data doesn't make it easier to modify a trained network.
0-_-0
·hace 9 días·discuss
If we can launch 1M satellites, how many telescopes can we launch?
0-_-0
·hace 11 días·discuss
32 bits vs 4 bits it looks like
0-_-0
·hace 12 días·discuss
Browser fingerprinting can already pinpoint you exactly. We should focus on that.
0-_-0
·hace 14 días·discuss
That math makes no sense
0-_-0
·hace 23 días·discuss
These never ever accomplish anything
0-_-0
·hace 25 días·discuss
Not if it's mass produced
0-_-0
·el mes pasado·discuss
No, it means the motor is smaller and it can be put into the wheel
0-_-0
·el mes pasado·discuss
And this is just fraud that was done with incompetence, so easily caught. How much is done competently?
0-_-0
·el mes pasado·discuss
This can't be used to save VRAM in practice. To generate a new token with the primary model, you first need to decompress the cache, which involves regenerating the whole sequence from scratch. I.e. generate 1 million tokens with the small model to generate 1 with the large.
0-_-0
·el mes pasado·discuss
You can use the original model to compress the kv cache and get ∞x compression, since the prediction is perfect. The cost is time, and I don't see how this could be worth it.
0-_-0
·el mes pasado·discuss
Agents don't buy stuff they see in an ad
0-_-0
·hace 2 meses·discuss
I thought it can do multiple tokens at a time