vrighter @ vrighter @discuss.tchncs.de

Posts

0

Comments

1,052

Joined

2 yr. ago

Any processor can run llms. The only issue is how fast, and how much ram it has access to. And you can trade the latter for disk space if you're willing to sacrifice even more speed.

If it can add, it can run any model

but we aren't any closer to agi than we were in the 50's. $100 billion in revenue for openai won't be any closer to agi either.

raytracing is insanely expensive. If you saw what current cards can render in real time, you would see a very very noisy, incomplete image that looks like shit. Without ai denoising and a lot of temporal shit (which only looks good in screenshots). It is very very very far from being able to render an actual frame with decent performance.

yes, but what you need to be doing is tons of multiply-accumulate, using a fuckton of memory bandwidth.. Which a gpu is designed for. You won't design anything much better with an fpga.

we use wired communications until the issue is solved.

it's actually unknown. It looks like it, but it is not proven

no, it isn't. A lot of systems still run power through the battery even after it's charged. The battery ends up in a state of constant trickle charging.

In the deck, once the battery is charged, and the power cable is connected, the battery is bypassed and effectively disconnected.

And the deck only charges to 100% if it was below 90 to begin with.

leaving it plugged in is perfectly fine, once the battery fills up, it is bypassed and the deck is powered directly off of the wire

re the shunt: yes it does

yes it is malware

i'd say it's closer to 105

played? Past tense? You must be joking

9% played 2024 games in 2023?

all the advertised AI detection tools are just that. Happy slapping!

because of the "perceptual" part.

A normal hash has the property that it produces wildly different hashes for even the tiniest of changes in the file.

Perceptual hashing flips that requirement on its head, and therefore makes finding a suitable hash function much harder.

This is not error correction issue though. Error correction means taking known data and adding redundancy to it so that damaed pieces can be repaired. This makes the message longer.

An llm's output does not contain error correction. It's just the output. And it doesn't contain any errors, mathematically speaking. The hallucination is the correct output. It is what the statistics it gathered from its training set determined is most likely. A "correct" llm output is indistinguishable from a "hallucination", mathematically, and always will be. A hallucination is simply "some output that some human, somewhere, doesn't like", and that's uncomputable. And outputs that people subjectively consider as "hallucinations" cannot be eliminated, because an llm is, fundamentally, a probabilistic algorithm. If you added error correction to an llm's output all you'd be able to recover is the llm's original output, "hallucinations" and all.

Tldr: "hallucinations" are a subjective thing. A Hallucination" is not an error that can be corrected after-the-fact, because it is not an error in the first place.

if you want to have a reliable driverless system, you're going to have to invent trains... again

they would need to develop balanced mechanics. Level scaling completely ruins any sense of progression.

they didn't. This is about a properly written headline by BBC being butchered when summarized by apple intelligence, which they have no control over.

the problems with (the current forms of generative) AI will not be solved, because they cannot be solved. They are intrinsic to the whole framework.