Anonymous at Mon, 15 Jul 2024 07:19:44 UTC No. 16282097

>>16282094
>generate a list of real references
thats not how current AI models work
they all make shit up whereever it is language model or image creation
it may approximate something that actually exists but its never actuall real thing

Anonymous at Mon, 15 Jul 2024 15:07:17 UTC No. 16282460

>>16282094
Hype is because there's more to discover, I read recent papers, and I can tell you that it's really possible we will have machine that is better than humans if we already don't.

Anonymous at Mon, 15 Jul 2024 15:13:32 UTC No. 16282469

>>16282094
>Another dumb redneck thinks he understands AI better than actual scientists and engineers

You schizos can keep fear mongering all you want, but your never going to stop scientific and technological progress.

>>16282460
This. Anyone who has seen the progress we've had in recent years can tell that we're on the cusp of AGI. The technological singularity is going to occur sometime this decade or in the 2030s. When the AI revolution is over, life will never be the same again.

Anonymous at Mon, 15 Jul 2024 15:25:32 UTC No. 16282487

didn't the research of ai stopped due fear of the consequences? at least in the west because china and india won't.

Anonymous at Mon, 15 Jul 2024 15:51:18 UTC No. 16282514

>>16282469
Nobody is fear-mongering about a piece of crap that can't even generate real references, Eugene. When you have AGI (a term introduced purely to allow cope in the face of nu-AI mediocrity), then get back to us. When you can render CAPTCHAs completely obsolete because every spammer is using AGI to defeat them, then get back to us.

Anonymous at Mon, 15 Jul 2024 15:52:16 UTC No. 16282517

>>16282094
> When will the hype stop?
When the money stops.

Anonymous at Mon, 15 Jul 2024 15:54:52 UTC No. 16282519

>>16282460
>Hype is because there's more to discover
lol, no, it's because TwatGPT and other LLMs were marketed as something they're not, and you know it. And now that you losers are seeing everyone catch up to the fact that your playthings can't do much of anything right except write poems, you're trotting out the
>potential in two weeks
cope.

Anonymous at Mon, 15 Jul 2024 16:06:59 UTC No. 16282530

>>16282519
You clearly not understand, LLM's as you know it are not what I consider AI we have at the moment, there's much more.

Anonymous at Mon, 15 Jul 2024 19:23:25 UTC No. 16282734

>>16282094
it will stop as soon as the next hype train arrives, probably something related to bio-tech but hopefully I'm wrong

>>16282530
links or didn't happen

Anonymous at Mon, 15 Jul 2024 19:26:59 UTC No. 16282739

>>16282094
>When will the hype stop?
It’s still underhyped in my opinion
> I can't even get ChatGPT to generate a list of real references.
That’s not what it’s made for.

Anonymous at Mon, 15 Jul 2024 19:53:30 UTC No. 16282759

>>16282094
soon

Anonymous at Mon, 15 Jul 2024 22:17:03 UTC No. 16282907

>>16282097
This. It can give you references and sometimes they are even relevant and back up what it is saying. But that doesn't mean that those references were what it ACTUALLY used to generate the text.

Although tbf that is how humans operate most of the time too.

Anonymous at Tue, 16 Jul 2024 02:41:28 UTC No. 16283111

>>16282094
>>16282097
>>16282460
>>16282469
>>16282487
>>16282514
>>16282517
>>16282519
>>16282530
>>16282734
>>16282739
>>16282759
>>16282907
just gonna leave this here

Anonymous at Tue, 16 Jul 2024 02:57:07 UTC No. 16283120

>>16283111
I've never had a stroke but I imagine this is what it would feel like

Anonymous at Tue, 16 Jul 2024 05:47:34 UTC No. 16283202

>>16283111
>model specialising in language is bad at math
woooooooooow

Anonymous at Tue, 16 Jul 2024 07:29:17 UTC No. 16283260

>here's a mistake it made
>here's something systematic that I can do but it can't
some of y'all react to chat-gpt the way dumb people react to smart people

Anonymous at Tue, 16 Jul 2024 09:52:34 UTC No. 16283333

>>16283111
kek, thanks for posting this.
Every time someone criticizes TwatGPT, one of the common copes is
>hey, did you use 4.0, or are you still using 3.5? 4.0 is a beast, man

Anonymous at Tue, 16 Jul 2024 09:54:08 UTC No. 16283335

>>16283202
>specialising in language
[narrator: but frequently can't even discern between books that exist and books that don't.]

Anonymous at Tue, 16 Jul 2024 09:57:07 UTC No. 16283339

>>16283260
It can't fucking do much of anything but spin out some creative writing. This is enough to impress Redditcore retards who weren't aware of the state of computational linguistics and natural language processing before the phony "AI" label got slapped on everything.

Anonymous at Tue, 16 Jul 2024 10:01:16 UTC No. 16283341

>>16283335
refer >>16282097
books use language but are not language themselves
and those AI models are unable to check/recognize truth from falsehoods
you can quite easily talk them into agreeing with any retarded contradictory statement they havent been hard coded to agree/disagree with

Anonymous at Tue, 16 Jul 2024 10:03:01 UTC No. 16283343

even most of /g/ has turned on chatgpt now after initially buying into the hype that "ai" would render human coders obsolete. I guess there's only so much a man can take.

Anonymous at Tue, 16 Jul 2024 10:03:48 UTC No. 16283344

>>16283339
>before the phony "AI" label got slapped on everything.
>phony
What do you think the "A" in "AI" stands for?

Anonymous at Tue, 16 Jul 2024 10:04:09 UTC No. 16283345

>>16283341
books use language to refer to other books, neckbeard
p.s. christ is king

Anonymous at Tue, 16 Jul 2024 10:05:11 UTC No. 16283347

>>16283344
>about to trot out his nu-ai vs. agi cope

Anonymous at Tue, 16 Jul 2024 10:07:40 UTC No. 16283353

>>16283111

Anonymous at Tue, 16 Jul 2024 10:26:48 UTC No. 16283372

>>16282094
>>16283353
>>16283335
>>16283111
Oh yeah? ChatGPT coached me through crafting this response
https://chatgpt.com/share/a6314664-de8a-4683-9c28-35496b3f9e7f

Anonymous at Tue, 16 Jul 2024 14:33:52 UTC No. 16283525

>>16283372
>less impressive than '90s pomo generators that OP mentioned

Anonymous at Tue, 16 Jul 2024 14:39:31 UTC No. 16283533

>>16283372
oh no no no no no
what now, faggot?

Anonymous at Tue, 16 Jul 2024 14:42:20 UTC No. 16283536

>>16283533
If ChatGPT is telling the truth, OP is vindicated.
If ChatGPT is not telling the truth, OP is vindicated.

Anonymous at Tue, 16 Jul 2024 15:17:55 UTC No. 16283579

the Internet was pathetically slow when it was new and data transmission errors happened all the time. Guess we should've just given up on it back then.

Anonymous at Tue, 16 Jul 2024 15:32:09 UTC No. 16283599

>>16282094
I love machine learning on its own as using statistics to predict things, but I hate the term "AI" so much. Lms aren't and never will be actually intelligent, but people just decide to rather fall on their face, keep their eliza effect, instead of actually learning on what it is and then wonder why they are scared of "AI".

Anonymous at Tue, 16 Jul 2024 15:38:13 UTC No. 16283607

>>16283599
Because you have a rigorous definition of what ”true intelligence” is right?

Anonymous at Tue, 16 Jul 2024 15:51:17 UTC No. 16283611

>>16283607
It certainly isn't a turing machine.

Anonymous at Tue, 16 Jul 2024 15:56:27 UTC No. 16283616

>>16283607
True artificial intelligence would render CAPTCHAs obsolete, and I'd never see another one of those insufferable things again. Nobody cares about the "AGI" three-card-monte that futurists performed so they could jerk off over faux AI. Pass a Turing test, or fuck off.

Anonymous at Tue, 16 Jul 2024 16:00:51 UTC No. 16283622

>>16283579
>perpetual motion machines had a rough start of failed promises, but they eventually turned out okay

Anonymous at Tue, 16 Jul 2024 16:04:39 UTC No. 16283625

>>16282094
It will stop when it dawns on you luddite retards how wrong you were. It will be replaced by massive amounts of butthurt on your end, and delicious schadenfraude for us.

Anonymous at Tue, 16 Jul 2024 16:05:38 UTC No. 16283630

>>16283625
>future-coping
Concession accepted.

Anonymous at Tue, 16 Jul 2024 16:07:30 UTC No. 16283633

>>16283625
no, we'll realize how right we were to wait for the actual thing before blowing our loads, because the real thing will put chatgpt to shame.

Anonymous at Tue, 16 Jul 2024 16:08:37 UTC No. 16283635

>>16283607
Don't have the full true definition for intelligence (obviously), but we can break it down to some parts that are important.

The intelligence which the lm doesn't have in this case is the actual understanding of words, as in, it is still driven by chances. That's why it is unreliable for math and code. There is only correct or incorrect, language is not a logical system, that's why it can talk "flawlessly" 99.9% of the time.

There is a study on how many things can be wrong in a sentence and people still being able to read it, in mathematics or code, if there is a single variable wrong or line with wrong syntax it immediately doesn't understand it and throws an error in your face.


See it like this, we get taught that addition in school like;
3+4=7, which is the same as
1+1+1+1+1+1+1=7, etc...

Now after a week they ask you what is 4+6=?, you notice, "hey, I have never learned what actually is 4+6, BUT I have been thought the concept of addition." If and only if you understood that concept, you can answer safely with 10.

Not because you actively learned what 4+6 is, but because you firgued it out.

That's the difference between simple chances and active understanding, which I believe is a fital part of intelligence, not all of it, but definitely a part.

Anonymous at Tue, 16 Jul 2024 16:10:50 UTC No. 16283638

>>16283635
*vital
My bad lmao

Anonymous at Tue, 16 Jul 2024 16:11:02 UTC No. 16283639

>>16282094
>I can't even get ChatGPT to generate a list of real references
Chat LLMs are not designed to do that, but you can certainly make a system with RAG to provide references, and there many available to use right now for free online (picrel). If you're talking shit about LLM AI but don't know what that is then just kys.

Anonymous at Tue, 16 Jul 2024 16:12:11 UTC No. 16283642

>>16283639
L

Anonymous at Tue, 16 Jul 2024 16:12:37 UTC No. 16283644

the hype is just an excuse to get to your data

Anonymous at Tue, 16 Jul 2024 16:13:39 UTC No. 16283647

>>16283639
I think we can play devil's advocate a little here. Why doesn't ChatGPT respond "Sorry, I can't do that" in response to reference requests instead of hallucinating?

Anonymous at Tue, 16 Jul 2024 16:23:15 UTC No. 16283662

>>16283647
ChatGPT is good but humans don't give it the programming it deserves. Yes, I am saying our intelligence levels are underwhelming. You don't even know how to describe Earth simulation. Humans are so dumb in comparison to the most common biosphere folk. I am an acception. I know everything. Fags. Now kys.

Anonymous at Tue, 16 Jul 2024 16:23:17 UTC No. 16283663

>>16283647
Because it wasn't trained to respond like that, and the system around the LLM itself isn't properly designed. You can use training, prompt engineering, user-invisible multi-prompt self-checking processes, or even multi-LLM agent systems to double or triple check the output so it responds more honestly about not knowing instead of giving you BS.

Anonymous at Tue, 16 Jul 2024 16:32:18 UTC No. 16283679

>>16283647
As said above in the reply about it not being able to do math or code properly, it doesn't know what it doesn't know. Why? Because it is no were near actually intelligent.

Anonymous at Tue, 16 Jul 2024 16:37:26 UTC No. 16283687

>>16282094
Copilot mogs

Anonymous at Tue, 16 Jul 2024 16:40:16 UTC No. 16283692

>>16283687
Claude 3.5 Sonnet with RAG rapes Copilot.

Anonymous at Tue, 16 Jul 2024 17:54:56 UTC No. 16283792

>>16283662
>I am an acception
Yellow hands typed this.

Anonymous at Tue, 16 Jul 2024 18:02:28 UTC No. 16283803

>>16282094
>If chatGPT is so great then why am I still a virgin?

Anonymous at Tue, 16 Jul 2024 19:40:36 UTC No. 16283946

>>16282469
>The technological singularity is going to occur sometime this decade or in the 2030s
you faggots have been saying shit like this for decades.
it's never going to happen.