"We invented a machine that thinks!"
"It's not thinking..."
"Technically true, but it can CREATE new content out of thin air."
"It steals everything it spits out by scraping the internet..."
"Okay, got me there, but think of the information access; now everyone with a phone has their own research assistant and the Library of Alexandria!"
"The machine spits out wrong answers; like every time."
"Look, fuck you, we own everything and you're buying it."
1. It doesn't matter if it meets your definition of "REALLY TRULY THINKING". It already beats you at a wide range of tasks you thought of as absolutely intractable for AI five years ago.
2. Yes, it steals everything and recombines it in creative ways. Like humans do.
3. Yes, it makes up stuff and that is a part of the reason why it's dangerous.
...for some notion of "beats" that doesn't involve being Actually Better at something.
@wrog @AnarchoNinaWrites Well, if you've lost a game of chess to an AI, for example, you can pat yourself on a back saying it's not ACTUALLY better than you at chess, it's just pretending somehow. But you still lost.
And if AI will kill you in various ways you couldn't even think about - it doesn't matter if it was ACTUALLY thinking. You'd still be dead.
@wrog @AnarchoNinaWrites Hmmm.
Well... Okay, so
1. ChatGPT kinda can play chess. Not very good though. But it's not because it can't in principle, but because it's not meant for playing chess. Nobody trained it to.
2. And you know why? BECAUSE WE ALREADY HAVE MUCH LESS ADVANCED AI'S THAT TRAIN 30 MINUTES FROM SCRATCH WITHOUT ANY INPUT OR FEEDBACK AND THEN EASILY SMOKE HUMAN CHAMPIONS, in pretty much any game, like for decades.
@wrog @AnarchoNinaWrites 3. ChatGPT sure can't play chess good enough. Yet. But you know what it can do? WRITE A FUCKING ACTUALLY WORKABLE CODE FOR AI THAT CAN.
4. So if I understood you correctly, if tomorrow they decide to train ChatGPT little more on games like chess and it suddenly can play it reasonably well - you will agree that it is dangerous? Or you'll come up with excuses like that right until you're dead?
1. Citation please
2. Citation please
3. Citation please
(It's like I'm actually arguing with ChatGPT; N.B.: making shit up is not an argument)
4. Who knows? We're not even close to there, and there's no indication we will be any time soon.
For now I have enough to worry about with the programs that actually ARE killing people in real life today (e.g., Tesla's "self driving" mode) and it's not due to them being "too intelligent".
1 and 2. reputable sources, please. I have no idea who the fuck "androidauthority.com" (a site I've never heard of before today) is and you probably don't either.
Also the statement "The world already has several chess-optimized AI that can outperform just about any human" is highly misleading.
ALL of the main contenders for best chess program are programs written by humans using deterministic algorithms. If you're calling those "AI" then you may as well call *all* software "AI". (yes, chess programs WERE, at one time, the focus of AI research, but that was 40 years ago. Researchers eventually figured out they weren't actually learning a whole lot about "intelligence" and moved on to other topics.)
3. Yes ChatGPT can search its data and find where someone has posted a chess-playing library and has scraped the documentation for it.
That's not a remarkable accomplishment. It didn't write the library, and if no one anywhere had posted a chess library it would have been up shit creek.
Nor is there any indication that the interface provided in your screenshot is actually correct (often what one finds with these things is you have to correct various mistakes before it'll run) but even if it is, there's no indication it didn't copy it straight out of the posted documentation. Not a whole lot of actual intelligence (in any sense of the word) here.
@wrog @AnarchoNinaWrites It can't "search its data", it's not how it works, it's not google. You can watch this after all the educational videos I sent earlier:
@wrog @AnarchoNinaWrites Oh come on, just calculate how many branches will it take to play chess with deterministic algorithm. Please educate yourself and don't talk about things you have no idea about. You can start with this:
https://youtube.com/playlist?list=PLZHQObOWTQDNU6R1_67000Dx_ZCJB-3pi
@wrog @AnarchoNinaWrites And maybe continue with this:
https://youtube.com/playlist?list=PLqL14ZxTTA4fRMts7Af2G8t4Rp17e8MdS
"It can't "search its data", it's not how it works,"
actually that's pretty much exactly how it works (except that the "data" is a collection of word-sequence frequencies based on the raw data it trained on). You need to learn about large language models.
Here, have an actual research paper
https://dl.acm.org/doi/10.1145/3442188.3445922
or, if you like, the Wikipedia summary:
https://en.wikipedia.org/wiki/Stochastic_parrot
which is unfortunately short but has links to lots of other stuff
Read all of the words, including, e.g., the 2nd sentence of the Wiki page I just gave you:
"are impressive in their ability to generate realistic-sounding language but ultimately do not truly understand the meaning of the language they are processing."[2]
and maybe re-read some of my previous replies with this in mind.
I'm also about this close -><- to blocking you.
"Please educate yourself"
fuck you. I already have a PhD in Computer Science from Stanford University, with specialties in algorithms and complexity theory.
YOU are the one who doesn't know what you're talking about, and I don't really care how many youtube videos you can pile up .. whether they're by idiots or you're misunderstanding what they're saying, it's not worth my time to go through them.
I've already tried to point out where you're going wrong but if you're not getting it, I'm done wasting time on you.
AnarchoNinaWrites
Бджілка робе кусь
Roger Crew✅❌☑🗸❎✖✓✔
