Posts: 1,616
Threads: 102
Joined: Aug 2007
Posts: 1,057
Threads: 121
Joined: Dec 2010
12/24/2022, 10:08 AM
(This post was last modified: 12/24/2022, 10:09 AM by JmsNxn.)
LMAO!
Bo, I have seen enough experiments with ChatGPT to know it's stupid af. Ask it to make specific list sorting; ask it to make "1/2 Zeta function calculator". It's totally garbage.
You could probably tell it:
$$
\text{tet}_K(z) = e^z\\
$$
Tell it; that's tetration. And now \(a\) tetrated to the \(z\) is just \(a^z\).
Fucking garbage tool. AI could never write Sheldon's program.
This is designed to replace useless programmers who make HTML documents all day, lmfao.
LET'S GO!!!!!!!
Posts: 367
Threads: 28
Joined: May 2013
MSE MphLee
Mother Law \((\sigma+1)0=\sigma (\sigma+1)\)
S Law \(\bigcirc_f^{\lambda}\square_f^{\lambda^+}(g)=\square_g^{\lambda}\bigcirc_g^{\lambda^+}(f)\)
Posts: 1,057
Threads: 121
Joined: Dec 2010
12/26/2022, 03:36 AM
(This post was last modified: 12/26/2022, 03:40 AM by JmsNxn.)
(12/24/2022, 10:27 AM)MphLee Wrote: Ai can do something maybe
https://www.quantamagazine.org/aireveal...20221123/
I find these types of articles pretty sensationalist. There are strong bounds on the Fast Fourier Transform, and various Matrix algorithms. If anything, this is just a fancy calculator that might be able to "guess" a better solution. But once you've bounded the speed of successive matrix multiplicationsand you've found an optimal solution; there's nothing A.I. can do. This is just computer science fluff; where computer scientists are running wild trying to increase speed. In reality; for perhaps, strange processes the matrix speed up will be viable. But no one uses these matrix processes. This is another "result" proven by code hungry Machine Learning Tech Bros; where the priority is actually given to some guy in the 60s who did it by hand, lmao.
But, I hate tech magazines. My neighbor reads all these tech magazines, and talks to me about it "because I'm the math guy". But the stuff's so stupid. I had to fully explain the right hand rule to him; because he thought that a bicycle traveling forward naturally falls to the right; and that was the right hand rule. So just know, that's their audience. Lmao. I had to carefully explain to him that it has to do with gyroscopes and how the momentum forward, and the spinning wheel, interact, where small lateral motions force a stronger force. Guess why bicycles fall to the right; because the gears are on the right and therefore add more gravitational force; what I mean by lateral force.
Posts: 1,742
Threads: 382
Joined: Feb 2009
(12/26/2022, 03:36 AM)JmsNxn Wrote: (12/24/2022, 10:27 AM)MphLee Wrote: Ai can do something maybe
https://www.quantamagazine.org/aireveal...20221123/
I find these types of articles pretty sensationalist. There are strong bounds on the Fast Fourier Transform, and various Matrix algorithms. If anything, this is just a fancy calculator that might be able to "guess" a better solution. But once you've bounded the speed of successive matrix multiplicationsand you've found an optimal solution; there's nothing A.I. can do. This is just computer science fluff; where computer scientists are running wild trying to increase speed. In reality; for perhaps, strange processes the matrix speed up will be viable. But no one uses these matrix processes. This is another "result" proven by code hungry Machine Learning Tech Bros; where the priority is actually given to some guy in the 60s who did it by hand, lmao.
But, I hate tech magazines. My neighbor reads all these tech magazines, and talks to me about it "because I'm the math guy". But the stuff's so stupid. I had to fully explain the right hand rule to him; because he thought that a bicycle traveling forward naturally falls to the right; and that was the right hand rule. So just know, that's their audience. Lmao. I had to carefully explain to him that it has to do with gyroscopes and how the momentum forward, and the spinning wheel, interact, where small lateral motions force a stronger force. Guess why bicycles fall to the right; because the gears are on the right and therefore add more gravitational force; what I mean by lateral force.
yeah !
those magazines can be sh*t
more scifi and bs usually.
time travel and such
Posts: 1,742
Threads: 382
Joined: Feb 2009
(12/31/2022, 12:03 AM)tommy1729 Wrote: (12/26/2022, 03:36 AM)JmsNxn Wrote: (12/24/2022, 10:27 AM)MphLee Wrote: Ai can do something maybe
https://www.quantamagazine.org/aireveal...20221123/
I find these types of articles pretty sensationalist. There are strong bounds on the Fast Fourier Transform, and various Matrix algorithms. If anything, this is just a fancy calculator that might be able to "guess" a better solution. But once you've bounded the speed of successive matrix multiplicationsand you've found an optimal solution; there's nothing A.I. can do. This is just computer science fluff; where computer scientists are running wild trying to increase speed. In reality; for perhaps, strange processes the matrix speed up will be viable. But no one uses these matrix processes. This is another "result" proven by code hungry Machine Learning Tech Bros; where the priority is actually given to some guy in the 60s who did it by hand, lmao.
But, I hate tech magazines. My neighbor reads all these tech magazines, and talks to me about it "because I'm the math guy". But the stuff's so stupid. I had to fully explain the right hand rule to him; because he thought that a bicycle traveling forward naturally falls to the right; and that was the right hand rule. So just know, that's their audience. Lmao. I had to carefully explain to him that it has to do with gyroscopes and how the momentum forward, and the spinning wheel, interact, where small lateral motions force a stronger force. Guess why bicycles fall to the right; because the gears are on the right and therefore add more gravitational force; what I mean by lateral force.
yeah !
those magazines can be sh*t
more scifi and bs usually.
time travel and such
That comment applies more to some science mags i guess.
I do like the advancement made by AI for the matrix multiplication.
Me and my friend Mick have considered fast mulitiplication algoritms for various things such as matrices , 8 dim numbers , polynomials etc.
So call me biased lol.
See for instance :
https://math.stackexchange.com/questions...81#1888681
Since matrix multiplication occurs here too it might sometimes help.
regards
tommy1729
Posts: 1,057
Threads: 121
Joined: Dec 2010
01/23/2023, 03:50 AM
(This post was last modified: 01/23/2023, 03:54 AM by JmsNxn.)
(01/01/2023, 01:47 PM)tommy1729 Wrote: (12/31/2022, 12:03 AM)tommy1729 Wrote: (12/26/2022, 03:36 AM)JmsNxn Wrote: (12/24/2022, 10:27 AM)MphLee Wrote: Ai can do something maybe
https://www.quantamagazine.org/aireveal...20221123/
I find these types of articles pretty sensationalist. There are strong bounds on the Fast Fourier Transform, and various Matrix algorithms. If anything, this is just a fancy calculator that might be able to "guess" a better solution. But once you've bounded the speed of successive matrix multiplicationsand you've found an optimal solution; there's nothing A.I. can do. This is just computer science fluff; where computer scientists are running wild trying to increase speed. In reality; for perhaps, strange processes the matrix speed up will be viable. But no one uses these matrix processes. This is another "result" proven by code hungry Machine Learning Tech Bros; where the priority is actually given to some guy in the 60s who did it by hand, lmao.
But, I hate tech magazines. My neighbor reads all these tech magazines, and talks to me about it "because I'm the math guy". But the stuff's so stupid. I had to fully explain the right hand rule to him; because he thought that a bicycle traveling forward naturally falls to the right; and that was the right hand rule. So just know, that's their audience. Lmao. I had to carefully explain to him that it has to do with gyroscopes and how the momentum forward, and the spinning wheel, interact, where small lateral motions force a stronger force. Guess why bicycles fall to the right; because the gears are on the right and therefore add more gravitational force; what I mean by lateral force.
yeah !
those magazines can be sh*t
more scifi and bs usually.
time travel and such
That comment applies more to some science mags i guess.
I do like the advancement made by AI for the matrix multiplication.
Me and my friend Mick have considered fast mulitiplication algoritms for various things such as matrices , 8 dim numbers , polynomials etc.
So call me biased lol.
See for instance :
https://math.stackexchange.com/questions...81#1888681
Since matrix multiplication occurs here too it might sometimes help.
regards
tommy1729
Yes, this is my point tommy.
Many "specific cases of matrix multiplication" can be sped up using A.I. Algorithms. But ultimately, the most general constructions are proven and designed to be the most optimal. Matrices are such vast ideas that, you can throw so many wrenches in the gears, that the top algorithms go slower. So if you "throw a wrench in the gear", and tell an A.I. to find the "fastest algorithm"you'll definitely get some advancements.
But I was recently reading, before I posted this, a short article by a mathematician/machine learning person. And he said so fucking many of these "speed ups" discovered by Computer Scientists, are actually results proven in the 60s/70s. That they are rediscovering. They are just simply rediscovering them in a broad sense of "machine learning". But machine learning is just linear optimization. And these computer scientists cannot understand the mathematics, and are, in effect, rediscovering these results in a different language. But they are standard mathematically derived (without a computer, mind you) linear optimization techniques.
They are instead, running 1000 iterations of a program to approximate the linear optimization that had been shown in the 60s/70s. But they are saying "only machine learning can accomplish this". When really, it's attracting towards the optimal solution, that was already known and discovered by hand. Think, Kneser figured out tetration by hand with no calculator, and these machine learning tech bros think they discovered Kneser by running some matrix approximation program. And that the media talks about the tech bros, but don't talk about Kneser doing this shit by hand
Honestly, as a mathematician, I hate machine learning. It's basically just multiply a bunch of matrices together to get an "intelligent" thing. It's stupid af. Real intelligence is far more adaptive than this garbage. Proof is "machine intelligence" rediscovering matrix speed ups that people did by hand....
And the quip I always use, intelligence isn't:
$$
\prod_{j=1}^\infty A_j\\
$$
Which is pretty much 99% of machine learning. Multiplying a bunch of matrices acting on a vector space is not intelligence. It's just so wrong....
Posts: 1,742
Threads: 382
Joined: Feb 2009
(01/23/2023, 03:50 AM)JmsNxn Wrote: (01/01/2023, 01:47 PM)tommy1729 Wrote: (12/31/2022, 12:03 AM)tommy1729 Wrote: (12/26/2022, 03:36 AM)JmsNxn Wrote: (12/24/2022, 10:27 AM)MphLee Wrote: Ai can do something maybe
https://www.quantamagazine.org/aireveal...20221123/
I find these types of articles pretty sensationalist. There are strong bounds on the Fast Fourier Transform, and various Matrix algorithms. If anything, this is just a fancy calculator that might be able to "guess" a better solution. But once you've bounded the speed of successive matrix multiplicationsand you've found an optimal solution; there's nothing A.I. can do. This is just computer science fluff; where computer scientists are running wild trying to increase speed. In reality; for perhaps, strange processes the matrix speed up will be viable. But no one uses these matrix processes. This is another "result" proven by code hungry Machine Learning Tech Bros; where the priority is actually given to some guy in the 60s who did it by hand, lmao.
But, I hate tech magazines. My neighbor reads all these tech magazines, and talks to me about it "because I'm the math guy". But the stuff's so stupid. I had to fully explain the right hand rule to him; because he thought that a bicycle traveling forward naturally falls to the right; and that was the right hand rule. So just know, that's their audience. Lmao. I had to carefully explain to him that it has to do with gyroscopes and how the momentum forward, and the spinning wheel, interact, where small lateral motions force a stronger force. Guess why bicycles fall to the right; because the gears are on the right and therefore add more gravitational force; what I mean by lateral force.
yeah !
those magazines can be sh*t
more scifi and bs usually.
time travel and such
That comment applies more to some science mags i guess.
I do like the advancement made by AI for the matrix multiplication.
Me and my friend Mick have considered fast mulitiplication algoritms for various things such as matrices , 8 dim numbers , polynomials etc.
So call me biased lol.
See for instance :
https://math.stackexchange.com/questions...81#1888681
Since matrix multiplication occurs here too it might sometimes help.
regards
tommy1729
Yes, this is my point tommy.
Many "specific cases of matrix multiplication" can be sped up using A.I. Algorithms. But ultimately, the most general constructions are proven and designed to be the most optimal. Matrices are such vast ideas that, you can throw so many wrenches in the gears, that the top algorithms go slower. So if you "throw a wrench in the gear", and tell an A.I. to find the "fastest algorithm"you'll definitely get some advancements.
But I was recently reading, before I posted this, a short article by a mathematician/machine learning person. And he said so fucking many of these "speed ups" discovered by Computer Scientists, are actually results proven in the 60s/70s. That they are rediscovering. They are just simply rediscovering them in a broad sense of "machine learning". But machine learning is just linear optimization. And these computer scientists cannot understand the mathematics, and are, in effect, rediscovering these results in a different language. But they are standard mathematically derived (without a computer, mind you) linear optimization techniques.
They are instead, running 1000 iterations of a program to approximate the linear optimization that had been shown in the 60s/70s. But they are saying "only machine learning can accomplish this". When really, it's attracting towards the optimal solution, that was already known and discovered by hand. Think, Kneser figured out tetration by hand with no calculator, and these machine learning tech bros think they discovered Kneser by running some matrix approximation program. And that the media talks about the tech bros, but don't talk about Kneser doing this shit by hand
Honestly, as a mathematician, I hate machine learning. It's basically just multiply a bunch of matrices together to get an "intelligent" thing. It's stupid af. Real intelligence is far more adaptive than this garbage. Proof is "machine intelligence" rediscovering matrix speed ups that people did by hand....
And the quip I always use, intelligence isn't:
$$
\prod_{j=1}^\infty A_j\\
$$
Which is pretty much 99% of machine learning. Multiplying a bunch of matrices acting on a vector space is not intelligence. It's just so wrong....
I asked w alpha something about the riemann zeta function.
He drew me a circle.
lmao.
sad and funny.
point is all intelligence essentially comes from humans and luck.
mathematicians , good programmers , detecting patterns and getting lucky.
Im not a fan of AI at the moment.
chess engine is all i need lol
well i do use some software occasionally but usually dissapointed.
Getting a number does not imply " understanding " anyway.
regards
tommy1729
