Science

Jim__

(14,576 posts)

3. I'm not sure why it would affect computational number theory and encryption.

Sun Oct 13, 2024, 05:44 PM

Oct 2024

My reading is not that they are changing any underlying algorithms, they are changing the implementation of the algorithms. Apparently AI algorithms perform an extremely high number of floating point multiplications. They are converting those floating point multiplications to integer addition, and that uses significantly less electricity.

The computations are less energy intensive. I don't believe they are any faster or any more powerful than the computations they are replacing.

Again that's based on my reading, you may be seeing something different.

Edit history

Please sign in to view edit histories.

Recommendations

0 members have recommended this reply (displayed in chronological order):

7 replies

= new reply since forum marked as read

Highlight:

Integer addition algorithm could reduce energy needs of AI by 95% [View all] Jim__ Oct 2024 OP

Thanks. littlemissmartypants Oct 2024 #1

Screw AI -- how does this affect computational number theory and encryption ? eppur_se_muova Oct 2024 #2

I'm not sure why it would affect computational number theory and encryption. Jim__ Oct 2024 #3

Computational number theory does a great many giant-integer multiplies. The fastest way to do these ... eppur_se_muova Oct 2024 #4

It sounds like they may be obverse problems. Jim__ Oct 2024 #5

Encryption surely requires exact integer arithmetic, rather than approximate floating-point arithmetic muriel_volestrangler Oct 2024 #6

It reminds me of tricks that were used to get 3D gaming effects on slow 8 and 16 bit computers. hunter Oct 2024 #7