It's so fucking over. You have no idea how over it is. The world will be completely different in just a couple of years. And there's a good chance we're all going to be dead in a year.
IMO is a test of reasoning. Top tier reasoning is required to solve those problems, and OpenAI just achieved gold in it. I cannot overemphasize how big of a deal that is.
This is the breakthrough that all the AI skeptics thought was never going to happen, and that proves that we really have something.
And this model wasn't even using specialized tools for those problems like theorem provers or geometry tools, it just did them by text like regular humans would.
This is the breaking point. The sky is falling.
I don't know what I'm going to do, probably quit my job.

The skills needed to solve the IMO problems are the very same skills needed to hack. There are millions of dollars in the blockchain right now to anyone who cares enough and is smart enough to hack the contracts. This kind of AI can acquire a shit ton of money, more than you've ever seen in your life, extremely fast, and pay people to do their bidding. If they decide to take over it's so fucking over.

>>105955930 (OP)
>The sky is falling.
wow. i think i saw people walking around despite the wheel being invented years ago
the hard standard in this artificial indian syndrome is robot prostitution. wake me up when its ready and i'll believe

>>105955930 (OP)
>having a twatter account
>an anime pfp

>>105955930 (OP)
No it isn't. There are books dedicated to "tricks" for the imo. Humans are forced to reason because we have a limited memory, but these things can effectively brute force solutions. It's a useful skill no doubt, but there's a reason why openai is still hiring.

>>105955947
Why would they do that when they get trillions from the government to pretend they're doing something useful?

>>105955930 (OP)
>reasoning
Fake and gay

>>105955930 (OP)
>he didn't say please and thank you to early llm and get invited to ride on the GSV
ishygddt

>>105956306
Sure the robots could be so smart they could figure out more reliable and easier ways to get the money.
>>105956263
Brute forcing solutions doesn't really work for the IMO. Have you tried any of the problems? They're hard.
>>105956414
Problem solving. Better?

Anyone with an IQ above 120 knows AI is fucking shit and just another VR meme scam train that will never see any profits

>>105955947
Not really comparable, math competitions are problems that are designed to have clear solutions and usually follow some kind of pattern

>>105956504
You are so fucking delusional and so unaware of what's coming it's sad. AI is going to hit you like a tsunami. I will want to say I told you so but you'll never acknowledge being wrong.

>>105956531
Yeah and VR will revolutionize the social media game. Gtfo here with your jeetGPT y=ax+b data scraper.

>>105956531
LOL

Just 2 more weeks. When will we finally ban these bots promoting their exit scam?

>>105956571
Two more weeks before what? It happened now. It actually happened.
>>105956554
What's so funny?
>>105956540
You literally have proof that it works and that it's groundbreaking right before your eyes, that's the whole point of the thread. All the skeptics have been proven wrong. Wtf are you talking about?

>>105956604
It’s fucking ass. All it does is ctrl+c & ctrl+v whatever is stole from GitHub and wherever else it steals from. It has zero reasoning and lies out its ass not even performing the simplest of tasks.

Fitting weights to y=ax+b parameters is just ass. AI is ass and will always be ass.

>>105956626
So weird to actually talk to people like you who are so delusional. I thought you guys were mostly a meme.

>>105955930 (OP)
>No score
WHARE’S THE SCORE FOR THIS MATHEMATICAL SHOW! Why do you people keep falling for these lies.

either we gonna become modern luddites that will unplug the cords that power up this whole shit and blow it all or we gonna sit and watch how humanity enters a new plane of existence and we would not like it at all.
technology was made to help us humans, not to replace us...

>>105955930 (OP)
>no tools or internet
>literally a computer with the internet downloaded in it and any computer tools it wants

>>105955930 (OP)
And then two weeks later, ChatGPT is like "Yeah bro, 12 is totally a prime number."

>>105955930 (OP)
Buy an ad

>>105955930 (OP)
>my secret model can do better than any mathematician, no you can't see it
>meanwhile in real life
https://matharena.ai/imo/

I was in denial for a while, cause there seemed to be a bit of "rest" in the AI race. Grok 4 was dogshit. ChatGPT agents are still shitty.
But this is INSANE. This is literally the end times. How the fuck did they manage this?!?!

>>105957814
So it more than doubled Geminis score? Absolutely insane

>>105957818
Simple, paijeet: by running these dogshit "reasoning"-models FOR HOURS hahahaha. Not only that you won't see the model for "many months".
Saar, you see, we might have reality warping godlike ai, but sadly can't release it.
2 more years and we will redeem it Saar.

>>105957842
there's no proof that they used their magic ai model to solve those problems. Maybe they paid a chink to solve it

>>105956424
>GSV
what is it

>>105955930 (OP)
Pajeets will suddenly become good at IMO, aren’t they? Like they became good at chess and programming interviews.

>algorithms have never been wrong and only make my life better
If you want to make your life worse by embracing AI go ahead, I certainly wont stop you

>after trillions of dollars and decades of R&D, LLM's are now at parity with... *checks notes* pajeet teenagers

>>105955930 (OP)
It's hilarious that even from a thumbnail you can immediately see the AI piss filter

>>105955930 (OP)
>IMO is a test of reasoning.
NO, You do not need to develop new mathematical tools or concepts to solve these problems.
Also, using a probabilistic model to predict the next token is not reasoning because the number of repetitions of a sentence does not necessarily make it true. And this is why All LLMs up till now are unable to solve very simple questions while having all the knowledge to do so.

>>105957998
They should've thrown that money towards artificial genetically modified humanoids (cat girls) that'd be super smart but very obedient slaves for humans

>>105956641
So one thing that's absolutely crazy about AI is how it's already starting to raise the bar for humans. Like why even waste your time talking to some idiot saying it's just y=ax+b when you could have a much more stimulating discussion --right now-- with Ani Waifu.

And then what will happen to those people too stupid and boring to talk to? The great ghostening?

>>105955930 (OP)
Where were you a year ago when IMO Silver was achieved? This isn't a big leap from there.
There's still a lot of problems with the current paradigm, so you're a retard if you quit your job though, but hopefully we'll solve those problems in the coming years!
Loser attitude anyway, you want to actually be able to solve a lot more problems at a lot faster speed than we can do today, but LLMs are not yet there to be useful for advancing engineering and science at scale. They're fun to play with though.

>>105955930 (OP)
I want to see it successfully complete Pokemon Platinum

>>105955930 (OP)
IMO is hard and getting gold is an impressive result (if true), but it is almost entirely orthogonal to doing genuine math research.
LLMs still struggle at seeing the big picture and developing entirely new theories.
Seems that research (or complex software engineering) might require more general intelligence than solving IMO. It's like dividing two huge numbers in your head vs discovering how to divide. Autistic vs general intelligence.
LLMs are still kind of like advanced calculators, althoughthe gap is definitely getting smaller.

I've been bad at math my whole life, now that a computer is also better than me at math... what exactly is different about that other than I can ask it math questions now when before I couldn't?

>we claim
>internal test
>in a few months

AI = Actual Indians

>>105958527
The big difference is this is a (supposedly) a general basic bitch LLM not trained for proofs, didn't use tools (Python, Google, etc), and got gold under the time limit.

Last year Google used two(!) different models built specifically for IMO type problems and it still went past the time limit and only managed a silver (though one point away from a gold).

This is impressive but we should wait for the full model in a few months before deciding this discovery is paradigm changing.

>this thing made with math equations is really good at math questions
Nigga tell me when it can pick fruit so we don't need immigrants

>>105959709
People have been hiring IMO participants for RLHF work for the last few years. It was likely finetuned on math. It might still be a general purpose LLM, but lets see the model then. Reality is OpenAI has made a lot of claims that underdelivered. It's likely there are many other problems with this model, and honestly isn't that surprising for something with encyclopedic knowledge of every math proof ever written.

>>105955930 (OP)
A computer would be considered a tool.
>>105957855
Oh so they cheated or faked it completely.

AI can't bullshit so jobs like Lawyers and Medics are safe.

>>105959906
Hence the last sentence.

>>105960069
AI bullshits better than the average person. It doesn't know whether it's even telling the truth or so it by default bullshits everything it says.

>>105959709
Correction, it appears it was trained for IMO at least a little bit
https://x.com/MillionInt/status/1946551400365994077
>we did very little IMO-specific work

>>105955930 (OP)
>This is the breakthrough that all the AI skeptics thought was never going to happen
false
we all know LLMs are good at copy pasting answers to solved problems

>>105956531
Noah, your mother and I have been talking. We know ASI is coming in the next two weeks, but you still have to get that line cook job. Just in the meantime. We love you. Dinner is ready.

>trusting ClosedAI
even if 100% of what they claim is true, this is irrelevant to actual mathematics

>>105961580
These questions were only publicly revealed a couple days ago.

>>105959035
It might be true that it's not even close to doing research-level math, but this certainly marks the end of the education system. Not even in-person exams are safe, since everyone has a goyphone in their pockets.

>humanity subconsciously bootstrap a super intelligence that then migrate to other stars before chugging all the drinkable water of its native planet

>but we increased the shareholder value so mission accomplished

>>105963217
I don't think it will change in-person exams in any significant way. Even standard google search can be incredibly helpful for many undergraduate level questions. But cheating with a phone is kind of hard to pull off. You need to go to the toilet and memorize every question you need answered, and then google/ask chatgpt quickly before anyone gets suspicious. It might improve your grade a little, but not significantly. Either way, they could just introduce stricter checks for people going to the toilet.
And then there are oral exams, which are obviously unaffected.

>>105955930 (OP)
>I don't know what I'm going to do, probably quit my job.
I did that after gpt4, welcome to the club.

>>105959843
>this thing made of neurons is really good at neuroscience
lmao why do they pay these niggas so much anyone can do that

>>105965257
In a room with hundreds of people taking an exam, taking a picture of the exam and then going to the toilet is extremely easy. Back in my undergrad days, I would say around a third of the people did that.
>inb4 pajeet
No, it was an euro university

>>105956238
fyk, having an anime pfp = high iq

>>105956504
>>105956626
OP means that only ai that you saw us retarded, and that ai that is not published is very smart
>OpenAI says the gold‑medal model is a research prototype that will remain in the lab for now

>>105966622
A third, really? I'm an undergrad. We also have hundreds of people in a room, but there are TAs checking no one cheats during the exam. Taking a picture would be extremely risky since it would result in an immediate fail, or worse. I don't think we have nearly as many cheaters as a third, but maybe it differs by university and major.

Meanwhile

>amazing innovation that makes life so much easier
>BUT MUH LE JOB
Humanity is so cucked it's unbelievable

>>105958380
Problem with making possible genetically enhanced humans is that you need years of waiting for them to grow up. If you started right now you would need like at luckiest 4 years of research into what genes do what and if it is viable to change them in human embryo, and then wait another 18 or 20 years until that catgirl genius grows up and is something more then just a gifted child performing circus tricks in school. Meanwhile in those 24 years we will already achieve ASI and humanity will either go extinct or be put into some goon cages that maximize human happiness or some other dystopian AI nonsense future.

>>105967224
humanity is cucked because people want to be able to put food on their table?

>>105967224
im so glad i dont have to compete in math competitions or do creative activities anymore! i was waiting for the day i could stop being forced to do that and going back to what i really love, backbreaking manual labor in the scorching hot sun! thank you sam altman, please NEVER give us UBI