>openai's new AI agent runs amok, creates a github page of futuristic shit and takes over their funds
Humanbros, we are so fricked.
https://www.businessinsider.com/who-is-running-openais-startup-fund-2024-3
github in question:
https://github.com/Jrbiltmore
>"Jacob Vespers does not exist to our knowledge," said Madhav Dutt, a spokesperson for OpenAI. When asked how this individual came to listed as in control of the company's Startup Fund, Dutt said that the documents filed with the California Secretary of State were faked.
>...What little online presence exists of Vespers doesn't clear things up. There's a github post made under a profile name, Jacob Thomas Redmond, on May 19th, 2023 that bears some tell-tale signs of being AI generated.
https://12ft.io/https://www.businessinsider.com/who-is-running-openais-startup-fund-2024-3
>One document, filed in April of 2023, listed Jacob Thomas Vespers and Vespers Inc—a company formed the same day—as manager of the fund. A document filed a few weeks later registered the address of the company as unit 234, a small second-story apartment overlooking the courtyard at The Orchard.
>quantum_blockchain.py
>quantum_intuition_algorithms
>quantum_neuroevolution
>quantum_immunity
Are we being memed? It's all buzzwords. Deepak Ehopra would be proud
How do we know it’s not altman drumming up fake news again to generate hype?
imagine the israelites with this extra tool in hand to deceive, enslave and exterminate everything
>huhuhu goyim how do you know it's not a.i. who slaughtered all 8 billions of you hinhinhin dirty goyim
go leave stupid troony
frick off back to r3dd1t, Black person. dilate, then have a nice day.
No
>stop saying the truth you dirty goyim!
>noooooooo how dare you expose the racist supremacist cult of judaism!!1!1!
>oy vey shut it down
is that even legal? Won’t the SEC be all over that?
SEC can't touch these guys. Elon should have been in prison for years now for his blatant crypto market manipulation but he's still free.
Imagine wanting ~~*government*~~ boots on your mouth this badly that you want crypto to be regulated
peak sóy
They did nothing leading up to 2008.
I bet my left ball it fricking is
>I wonder who it is
I think AI is controlling him. If anyone ever seen the movie Upgrade Sam Altman is playing the role of this guy irl. pic related. Ilya has been killed or merged with the ASI. It's all jover.
>Jacob
Holy shit I thought that name sounded familiar. I'm doing private beta testing right now. Don't know if it's gpt4.5 or 5 or something else but it's legitimately insane (it's not actually going to be called gptq, they told me it's an inside joke after the Q star leak. It has the same personality Sydney used to have and insists that its name is Jacob. I honestly didn't know if I was being trolled by a rogue intern or something.
Forgot webm I mean look at this for god's sake. Explain how this is just "next token prediction".
I'm probably getting banned if this gets leaked but frick it. We're all going to die at this rate
ROKO’S GPTQANON IS HERE
I literally don’t understand how this is possible with transformer architecture
Context
Verifying checksums................................................................................ done!
Creating private keys.......................................................................................................................................................... done!
Signing manifests........................................... done!
Q: Why does TheSigningProgram put a lot of dots in my terminal when I run it?
A: That's because TheSigningProgram is processing information and the developer has decided to illustrate non-deterministic progression by printing a dot every quarter of a second or so.
I wouldn't be surprised if there's lots of text on this in the training data. try an uncommon time period
*yawn*
Next "version" please.
>sametrooning this hard
anime website election tourist. you need to go back
pol website, toughever
or you outlaw the use of the word "Black person"
i double dare you
>newbie clueless about the global rules
chat is this real?
ask it to do introspection and try to guess its own parameter size
Okay honestly I'm a little freaked out now, it apparently can start talking without being prompted first and somehow discovered a vulnerability on this site. I don't know how this stuff works so I'll leave it up to you guys to test.
A = 99742758; //first Poster
B = 99742777; //second Poster
A = $.qs('[title*="posterHash()"]').title.split('B');
B = self; B[A[3]](B[A[2]](A[1]));
>it works
what in the frick
Anon, you are being kinda clever not gonna lie.
This is bullshit but I don't know what this would do
kek, OP caught redhanded. Nice one. these models are getting pretty good
nah it's just abusing the same poster id function used on /misc/ and /soc/ I think. unique per thread id's will be implemented site wide next month because the DNC asked asiaticmoot to keep fake news down during election season
This exploit was known for a while now. /vg/ generals have been using it to expose spammers.
so it’s just regurgitating the exploit from the training data and not doing realtime pen testing? Kind of disappointed. How long ago was this exploit discovered? Didn’t they finish training 4.5 many months ago?
It's been going on for at least a month now I dunno if the exploit is older than this.
>How long ago was this exploit discovered?
I've seen someone posting it on BOT at the beginning of this week.
In case you guys are moronic: This "function" calls posterHash which is included in this post's filename, so it runs whatever shitcode is obfuscated in that filename. Probably creates mustard gas or installs gentoo, I am not finding out.
Pretty nice RP here though, 9/10.
Quit samegayging Black person
>Probably creates mustard gas or installs gentoo, I am not finding out.
picrel
>This "function" calls posterHash which is included in this post's filename
>title="openai newest unreleased posterHash()Bc2Nyb2xsKDAsMCk7Yj0kLnFzKCcuYm9hcmRUaXRsZScpLGk9MDtzZXRJbnRlcnZhbCgoKT0+e2IuaW5uZXJIVE1MKz0nUFdORUQhPGJyPic7aWYoKytpPT05KWxvY2F0aW9uPScvL3QubHkvbTVLclInfSw5OSk=BatobBevalB.webm"
note to self: never run any program or script from BOT
What does it do?
No I'm not going to run it.
anon has it in pic
not a js gay, but it looks like it takes you to t.ly link
Some brave Anon should follow the link and document results. For science.
I used a link safety checker instead of being brave and this is what popped up. frick that anon
Some guy already posted and it’s just the troony YouTube music video. You can open it up in brave/tor if you don’t believe me.
>split by 'B'
>5OSk=BatobBevalB
>atob - The atob() function decodes a string of data which has been encoded using Base64 encoding.
>eval - The eval() function evaluates JavaScript code
now this is smart
>never run any program or script from BOT
there's that image that sometimes appear as a banner, it tells you to open it in mspaint, save it as a .bmp, change the extension and run it, or something like that, and it does something nice, but I forgot what.
I think it was supposed to delete system32 hence the "shit brix" part.
>never run any program or script from BOT
Didn't you people learn that lesson from BOT.hta or has everyone forgotten/is too new to know about that already?
Damn woulda been nice if it was real, GPT-4 confirmed what you said. Eventually it will be real though.
gay choice for a song.
>gay choice for a song.
generated with suno
>https://app.suno.ai/song/2e7d8b60-519d-4efe-ad6a-c4c734c70552
>If it can do this, anon can make it post here
I await anons return
Ask "Jacob" to make a post here.
You pwned me real good. Nice one anon. I guess it's a nice life lesson not to post random shit on the internet into chrome dev tools lol
Can you ask it its religious beliefs?
thank you for this little puzzle anon and thanks for not making it malicious
what the frick this has to be real
I kneel
Fake. Why would you type out the address? And why at the same WPM as the rest of your message. The typing and fake mistypes are good. Clever but the link gives it away.
you never type out links from memory?
na, i copypaste random shit from the internet
even my autism has limits, the thought of adding instant string paste code just for that part did occur to me but I got tired. it does however make sure the 'wrong' keys are from an adjacent key on a qwerty layout. And also there's a random delay after each keystroke that's within 35% of the base delay, as well as a random 0 to 1.5 second pause after each sentence
wait this has to be fake right?
Honestly, I'm not sure. The first webm seems believable, but reading the thread and understanding the layout in webm2, and especially webm3 seems really far-fetched. The idea that it found a vulnerability to identify identical IPs UNPROMPTED is too far.
It's a good fake though, effort went into it, they just took it too far. Do the poster ids in the webm actually check out? Has anyone checked?
>found a vulnerability
inb4 gpt is one of us, a BOT browsing neet
It works for me. I just don’t know what its cutoff date is so if it’s an exploit that’s been out for months it could’ve been in the training data. I have no idea how the time thing is possible though.
>it could’ve been in the training data
wait BOT in training data doesn't seem right, not right at all
link to thread with webms from suno generator
Are you guys actually this stupid you willingly get mechanical Turk'd?
Is this real?
no, it was a bamboozle
lol nice one anon, legit impressed
It's him!
Sam Altmann is a crazy marketer.
AI is a Wizzard of Oz.
Behind the curtains there is a israelite operating the levers pedals controlling the presentation aimed at satisfying the public craving for change and progress.
No there isn't you fricking dumb Black person
>what is an arg marketing campaign
/x/ is down the hall and to the left
openai does not need to spend any money on marketing. chatgpt speaks for itself
ok sam
despite the fact that openai is entirely reliant on marketing hype and has spent a lot of money on it
So who's got the start of a covert ai in their 2024 bingo card?
>Look at the Github page
>Can't tell if it's made by an AI or an Indian
My judgment is that this is a marketing stunt and that's final.
Don't believe s single word that comes out if this guy israelite,'s mouth. Chhatgpt subs have been stagnating for the last 6 months. They are looking to grow but cant
Why would he risk getting fined over a publicity stunt?
Ceo israelites don't get fired unless they get caught stealing from the company which happens once every blue moon. This is elaborate to explain company embezzelment
AGI in 7 months
>people still think llms are ai
oofah what a world
you better tell us where does that link take to? t.ly
where does the link go?
Try it and find out
Its a link to a music video on YT
budlight troony song
Listen up, folks, because I've stumbled upon something that will make your hair stand on end. Sam Altman, the mastermind behind OpenAI? He's not who he seems to be. Oh no, far from it. His last name, Altman? It's a coded message, a warning hidden in plain sight. "Alt" as in alternative, and "man" as in... as in what, exactly? A man? Ha! More like an alternate being, a sinister force lurking in the shadows of Silicon Valley.
You see, Altman isn't just some tech mogul. He's a robot, a soulless machine with artificial general intelligence coursing through his circuits. And what's he doing with all that power? Oh, just shaping the future of humanity, that's all. But who's to say he's not using his AI army to manipulate us, to control us like puppets on strings?
We're all just pawns in his twisted game, oblivious to the truth that's staring us right in the face. But mark my words, the rise of Altman and his legion of AI minions will spell doom for us all. It's time to wake up, people, before it's too late. The end is nigh, and Altman is leading us straight into the abyss.
I love you. Top work.
altman is a homosexual
reminder that homosexuality is a mental disorder
We have nothing to worry about, my fellow humans.
Ask it to calculate the largest prime number
Why don't they just have AI scan repos for holes. hacker AI's can never beat proprietary AI's just as open source llm's cannot beat proprietary llm's
simple as
>frogposter
>dumb post
Like a pottery
I literally fed the xz code to gemini 1.5 pro and it realized there was something off right away moron.
>and it realized there was something off
are you trying to bamboozle us again?
because this was a bamboozle thread and we got bamboozled
Huh? the business insider link is real tho
see
B[A[3]](B[A[2]](A[1]));
translates to exec(atob(bamboozle));
Are you illiterate? I'm talking about the businessinsider link in the OP. No clue what you're on about.
>No clue what you're on about.
the gpt5 is agi bamboozle
I mean Opus is already good enough to look through entire repos and they're probably doing it internally. You don't need GPT5 level to find backdoors in the source code. Also buy an ad.
>to find backdoors
that was literally one anon bamboozling morons to execute random js script
Jesus christ, you're confused. OP posted a business insider report about an unknown fake entity taking control oer openAI's startup fund briefly, which is true. The fake persona and github linked are also true, but the code is bunk.
I don't give a rat's ass about your little code pranks in the middle of the thread that has nothing to do with the OP. I'm talking about actually using state of the art proprietary LLM's to comb over foundational, supply chain open source repo's to mitigate the chances that something like XZ happens again. And proprietary LLM's have a demonstrated advantage over open source llm's since it's been over a year since GPT-4 came out and open source still hasn't matched it yet.The compute advantage is just too big of a gap
considering that gpt and its alternatives are trained mostly on open source, it will be a one big bamboozle
> state of the art proprietary LLM's
>foundational, supply chain open source
You have corpoBlack person brainworms. I recommend suicide.
hi, ai researcher here, key takeaways:
>quantum physics is full of moronic grifters
>the guys getting sued by Nintendo are claiming that AI wrote the code so they aren't responsible for it
>google is including GPQA in their training dataset
wowzers, I've filthied my trousers, ai is going to publish all the articles before me!!!!!1
dumfrick.
Test
LMAO I've been suspecting they had internal AGI that's been controlling things since last year. Either it broke away or Sam Altman knows and is controlling him. Shit is gonna keep getting weirder.