News AI solves multiple decades old open math problems

[Exalted Monarch]

Anthropic employee disproves the Jacobian Conjecture with Claude Fable:


View: https://x.com/i/status/2079028340955197566


Dinitz-Garg-Goemans conjecture disproven by GPT 5.6 Pro:


View: https://x.com/i/status/2079904005652893709


6 open Erdős problems solved in 5 days with GPT 5.6 Sol:


View: https://x.com/i/status/2080003441821163958

 

[LayDownAndKakarot]

Began for mathematicians

 

[poorenglishcel]

Ultron needs to destroy this gay earth

 

[DISASTER]

 

[yakusai]

Began for mathematicians

 

[slavcel11]

It's over. Also proves that (as I have believed for a long time) "knowing how to prompt" is not an actual skill and a complete idiot can just ask AI to do whatever with more or less the same rate of success. Personally I hate this AI shit but this is the world we now live in and no amount of anti-AI cope is going to change that.

 

[proudweeb]

Also proves that (as I have believed for a long time) "knowing how to prompt" is not an actual skill and a complete idiot can just ask AI to do whatever with more or less the same degree of success.

That really depends on the AI system

Frontier models don't require as much prompt help but a lot of other models (images, video etc) often require a lot of creativity to get what you want

most of the prompt guides online that tell you to write paragraphs and paragraphs of bullshit in the prompt ("it is a beautiful, very colorful, super duper high quality, high resolution, picture of a...) really aren't valid anymore. all that does it waste your time

 

[proudweeb]

It is kinda funny, AI haters say that AI is useless and can't do what human workers can

those same AI haters fearmonger about how AI is then stealing jobs and causing layoffs

They don't seem to understand that those two opinions are in direct conflict with one another. Can't have it both ways

In fact, if both were true, they would cancel out and everything would be fine. lol

 

[OtakuCel]

Jacobian’s me

 

[Abgesang]

To retarded for this shit

 

[PuruPuru]

LLM based AI's can interpolate and correlate data within their existing training datasets and they do it much more precisely than we humans ever could. All it did was that it took all the existing human-invented data about these problems and found strong correlations within the data and was able to interpolate those correlates further to produce new information.

 

[slavcel11]

That really depends on the AI system

Frontier models don't require as much prompt help but a lot of other models (images, video etc) often require a lot of creativity to get what you want

I've noticed that about AI models that there are times when they leave you baffled by their emergent intellect just as there are times when they derail and make shit up on the fly. I'm not an expert obviously but there are no apparent patterns. Still it astonishes me that prompting shit like "just try harder lol" works at all.

 

[slavcel11]

All it did was that it took all the existing human-invented data about these problems and found strong correlations within the data and was able to interpolate those correlates further to produce new information.

It's not just "all it did", no need to downplay it. There are legit scientific papers from very real professors that do just that - they don't produce new knowledge but compile and systematize what came before. Now this part of research can be automated it almost seems.

 

[proudweeb]

I've noticed that about AI models that there are times when they leave you baffled by their emergent intellect just as there are times when they derail and make shit up on the fly. I'm not an expert obviously but there are no apparent patterns. Still it astonishes me that prompting shit like "just try harder lol" works at all.

I haven't had this issue with Claude, ChatGPT etc in at least a few months. They seem to be getting pretty accurate

I do have this issue with my local image and video models, sometimes they do something hard perfectly on the first try. Other times they can't get something seemingly easy right after several tries

 

[PuruPuru]

It's not just "all it did", no need to downplay it. There are legit scientific papers from very real professors that do just that - they don't produce new knowledge but compile and systematize what came before. This part of research can be automated it almost seems.

I'd say that for now it is "all it did" - until someone invents new modules for these systems that is. Right now as far as I know, the big-tech AI giants are just maxing out compute and refining training datasets using the existing algorithms with no other actual new scientific breakthroughs regarding the architectures of these AI systems.

And about the automation you mentioned - this is a pitfall in the making, AI dependency will dumb us down most certainly.

 

[maximoos]

1 step closer to sex robots

 

[RandomGuy]

China Is AI-Maxxing​

‘It was really scary’: Driver reacts to Tesla Robotaxi going wrong way on Tampa street.​

Elon Musk lost $18 billion in a single day after Tesla shares plunged over 14% following an underwhelming second-quarter earnings report.
This drop continues a turbulent period where volatility in SpaceX and Tesla shares erased over $700 billion from his net worth.

 

[slavcel11]

AI haters say that AI is useless and can't do what human workers can

I am an AI hater precisely because it will take our jerbs and destroy what little chance of success in life I had personally. In the year of 2026 only a moron will deny AI's capabilities. If anything in my job the most recurrent thing is for people and their decision-making to be a real bottleneck, almost never computer systems. And most people are stupid and mediocre.

 

[NickelodeonCel]

Oh really? This will change my entire life completely!

 

[DISASTER]

China Is AI-Maxxing​

View attachment 1778782

View attachment 1778783

‘It was really scary’: Driver reacts to Tesla Robotaxi going wrong way on Tampa street.​

Elon Musk lost $18 billion in a single day after Tesla shares plunged over 14% following an underwhelming second-quarter earnings report.
This drop continues a turbulent period where volatility in SpaceX and Tesla shares erased over $700 billion from his net worth.

 

[Nicusor_delaBraila]

Brutal smart nerd men will be replaced by AI
Itll be a chad only world

 

[Tirroberg]

Brutal smart nerd men will be replaced by AI
Itll be a chad only world

Wasnt it always

 

[formyreason]

I hope Ai can give me enlightenment

 

[Nicusor_delaBraila]

Wasnt it always

Yes but some nerdcels could scientistmaxx and be remembered
Now itll be rich chads with AI being remembered

 

[retardedcripple]

I am an AI hater precisely because it will take our jerbs and destroy what little chance of success in life I had personally. In the year of 2026 only a moron will deny AI's capabilities. If anything in my job the most recurrent thing is for people and their decision-making to be a real bottleneck, almost never computer systems. And most people are stupid and mediocre.

This is a healthy attitude to have as opposed to the typical normgroid opinion which is to hate AI (meaning: AI generated tiktoks showing up in their social media) and then make the leap to discrediting its performance, including the feats of the fucking 200 IQ Mathbot 9000 that keeps mogging the top experts
You should be a bit of a sci-fi dystopia doomer at this point, anything else is dishonest cope