Forgot your password?
typodupeerror

Comment Re:Take the loss, Cathay. (Score 2) 16

Relocating your headquarters does not unincorporate your incorporated entities within the jurisdiction in question, and that's the trouble they're in now. They have many onshore (Chinese) holdings.

An Irish corporation doing business in the US via a US-incorporated entity is subject to US law. The people within that parent corporation are very subject to US law if they step foot on US soil, and mostly subject to US law if they step foot on the soil of any country with an extradition treaty with the US.

Comment Re:Empathy??? (Score 1) 107

Don't flatter yourself

I'm not- you're doing that for me.

You think I don't see you trying to snipe people out of spite? I'm not fucking stalking you, you self important bufon. I'm making sure you don't come into my spaces to talk bullshit about other stuff you also don't fucking understand because the chip on your shoulder is so damn big...

Ya, you are.

Thankyou for again explaining that you got the words you're saying here from marketing from a company that stands to make money off you believing them. The fact you're relying on nobody involved understanding the contents of the page you posted was already taken into consideration.

What in the fuck are you talking about?
From the page:

A stochastic process has the Markov property if the conditional probability distribution of future states of the process (conditional on both past and present values) depends only upon the present state; that is, given the present, the future does not depend on the past.

Emphasis mine.

A Markov system's evolution depends only on the current state.
i.e.,
If the last token was A, the next token's probability stems only from that fact.
If that is not a true statement, then your system is not Markovian.

To demonstrate the difference, a Markovian token predictor would have a next state probability field of V^n, where V is the vocabulary, and n is how large your n-gram is.
A non-Markovian LLM has a next state probability field of V^c*V, where c is the configuration space of the context, so that a small model easily has multiple googols of probability space for its next token. They're literally not comparable.

One of those is simple and predictable. The other has a state space that's astronomical.
You can, if you really want to abuse the definition, say that an LLM is Markovian, in the same way that the universe is.
Again, feel free to educate yourself.
That is why it is stupid to compare an autocomplete with an LLM.

Comment Re:Empathy??? (Score 1) 107

What you did wasn't mockery- it was outright fucking buffoonery.
The argument you're implying is that models have self-execution abilities, and are in fact not calculated by programs.

By the way, any system that takes sequences of stochastic data is a markov system. Mathematically speaking. Please feel free to explain to me condescendingly that maths is secondary to computer science as if they're somehow in disagreement.

That is entirely incorrect, and I explained that to your stalker ass last time.
A Markov system must satisfy the Markov Property.
Feel free to educate yourself.
Maths and computer science aren't in disagreement at all, only reality and your ignorance.

LLMs are autoregressive- they consider the entire past history to predict the next token, not just the last token. They are not Markovian.
I'd love to have educated you in a less condescending manner, but honestly motherfucker- you earned the demeanor you got.

Comment Re:Empathy??? (Score 0) 107

How, precisely, do you imagine a bunch of vectors are applied to a series of text to create an image? Do you imagine that a little hermunculous within the machine rubs the vectors all over the text until an image grows forth from them like fungus?

Diffusion generated art is a process. The model is merely the weights used for that process.
It takes software to feed the noise and embeddings to the model in order to produce the image.

I swear to fuck you get dumber every time you try.

Comment Re:Black box not useful for artists (Score 1) 107

Bullshit. You're trying the mental gymnastics. I tried to gently tell you it was just going to make you look stupid.
You said:

And *all* would like their work to appear as *they* intended.

Which of course, no one ever denied. This is called a straw man.

It's not hard to grasp. Point me to one artist (who is not a burnt out husk) that would say "sure I don't care about the skin texture I labored to hard to make fucking pores and wrinkles for, just turn DLSS so you can see it better, as NVIDIA knows best". Give it a rest.

And this is you beating that straw man to a pulp.

All I need to do is point to one artist who judged his work by what the output of the model was, no differently than a complex filter in photoshop, which they also have no fucking clue what its technical details are, but a pretty fine grasp of what its non-technical details are- almost precisely analogous to how they'll use model-generated/augmented art.

You see, you aren't really trying to argue for their control. Because they have that. What you're trying to do is wrap up an anti-AI argument in a veneer of apologetics. It's transparent, and it smells of horse shit. Give it a rest.

Comment Re:Black box not useful for artists (Score 1) 107

First time I saw this in a paper. So, yeah, "artists", and you also seem to have no clue.

And you seem to have an elementary grasp of logic.

That an artist may want that proves my point, not yours.
Artists are not a homogeneous group. Some may want to bend the laws of physics. Others will be steadfastly against it.br. There is no amount of mental gymnastics you can execute to make your statement less ridiculous.

Comment Re:Empathy??? (Score 0) 107

I was talking about the games industry.

You don't think stable diffusion is used in games? That's silly- it's all over in them.

You know, the place that would use DLSS for their product. Well let's see if it gets adopted I suppose. And by whom.

Precisely. Either it will- or it won't.
It'll be a hit, or it will bomb. But either way, what it will not be, is contrary to the intent of those that put it in their game.

Comment Re:AI is not very intelligent and not improving. (Score 1) 153

If you're trying to imply that I think the Abrahamic biblical stories are an original work from the mind of god, then you're a fucking idiot.
If you think that the Abrahamic biblical stories are an original work from the mind of god, then you're a fucking idiot.

Ultimately- I suppose that means you're a fucking idiot.

Comment Re:AI is not very intelligent and not improving. (Score 1) 153

I did not compare "an LLM" with "an animal".
I compared a diagnostic criteria of an LLM with a diagnostic criteria of an animal.

If I say that a box is striped like a Zebra, am I bordering on the edge of psychotic?
Any inference you made on my regarding an LLM as "alive" was a projection on your part.

Slashdot Top Deals

Surprise your boss. Get to work on time.

Working...