[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math

View post   
File: 1.45 MB, 2048x1682, 04C2973D-87AB-47A4-8314-52BFC1EF7FCC.jpg [View same] [iqdb] [saucenao] [google]
15240170 No.15240170 [Reply] [Original]

We are possible a couple years away from AI that has robust mathematical intuition and is capable of developing a mathematical system that is eons more complete than status quo.
Decades of work might be produced by 3 second inference run.
How do mathematicians cope?

>> No.15240182
File: 335 KB, 3539x2119, 369_image.png [View same] [iqdb] [saucenao] [google]
15240182

Language models are a bad way of doing math. We need to keep working on formalizing all mathematics as something like pic related first.

>> No.15240212
File: 93 KB, 834x767, 1672717116899441.png [View same] [iqdb] [saucenao] [google]
15240212

>>15240170
Two more weeks.

>> No.15240219

>>15240182
>language is bad use this instead
>posts programming language syntax instead of something that isn't language based

>> No.15240226
File: 72 KB, 606x800, godel.jpg [View same] [iqdb] [saucenao] [google]
15240226

>>15240182
>We need to keep working on formalizing all mathematics
You're welcome to try

>> No.15240227

>>15240212
Post the one where it "proves" that 2 is irrational

>> No.15242771

>>15240170
We aren't. We haven't been anywhere close to it for decades. We still won't be. Computer mathematics will forever be what it always has been in Mathematica, Maple, etc.

>> No.15242987

i like how people in this board are completely un aware of Math and Science dedicated Ai that existed for years.

Already exist. Big Challenge today is not to create, it is to reach PhD level quality work by 2025. i am hopeful. And, if it does by 2025. Work it can do by 2030 will Amaze

>> No.15243002
File: 179 KB, 1514x1168, 20230302_103031.jpg [View same] [iqdb] [saucenao] [google]
15243002

>> No.15244435

>>15240212
>>15240227
https://pub.towardsai.net/meta-ai-new-model-can-solve-international-math-olympiad-level-problems-673dfb94f526?gi=5f8ab90da478

>> No.15244753

>>15242987
people here are dumb

>> No.15244776
File: 291 KB, 612x365, KL.png [View same] [iqdb] [saucenao] [google]
15244776

>year is 2030
>AI "proves" the Riemann hypothesis
>The proof is 5000 pages of indecipherable hieroglyphics
>Mathematicians have to admit the RH is proved but gain nothing from the proof
>Math is just the feeling on intellectual blue balls forever

>> No.15244777
File: 380 KB, 750x863, 1675026221330541.jpg [View same] [iqdb] [saucenao] [google]
15244777

>>15240170
>How do mathematicians cope?
By working on Pure Mathematics that computer architecture and programming cannont compute.

Thats how I cope, by doing work computers cannot.

>> No.15246842

bump

>> No.15246848

>>15244777
what do you mean by that? what parts of math do you think cannot be computed and are the "purest"?
I was under the impression that even highly abstract math can be easily implemented on modern computers. Isn't category theory used in programming languages? IIRC even mathematicians dismiss that as being "too abstract" and crazy

>> No.15246853

>>15243002
it's over

>> No.15247936

>>15240226
Gödel doesn't claim we can't formalize everything. He says that we can't proove everything. In fact, he incompleteness theorems rely on the fact that everything is formalizable so that Gödel Numbers can be produced.

>> No.15247985

>>15244777
Actually, pure mathematics these days is a lot of "symbology" and verbal deduction that LLMs excel at. That will be one of the first fields to fall.