General Knowledge
Region: AU
Wednesday 26 August 2026 15:45:29 GMT
64209
2239
38
92
Music
Download
Comments
..frame@@&$_corrup''d :
it's not "ai model that can't cross", it's simply a mathematical limit
2026-08-31 11:32:20
90
FJBO :
I believe that's because of how embeddings are weighted. LLMs are based on a statistical similitude not necessarily on actual reasoning yet.
2026-09-01 19:11:53
29
Maudfer :
This is proof AI can't reason. Means the efficiency of a model is still tied exclusively to the size/quality of the training data and consequentially the compute used to train it. The day a model crosses that line, means we (or the AI itself) has developed intelligence further than that provided in the training process. Means it evolves, it reasons new intelligence for itself.
2026-09-02 06:28:43
16
Sen :
You could call it the black wall
2026-08-27 11:32:21
27
Peter Wilson :
Can humans cross that boundary?
2026-09-01 06:00:39
3
djeda1234 :
you can't beat god
2026-09-01 21:44:14
0
Fahad :
Ask ai about it
2026-09-01 13:32:00
1
Patsy 0 :
what did you expect? a model that linearly reaches its lowest error rate and then abruptly becomes horizontal?
2026-09-03 14:21:52
1
Mark Markguyver Gibb :
🤷♂️We l8ve within the confines and rules of our own simulation. Safety protocols dictate that simulations inside the simulation can only be up to a certain level of complexity before they are automatically stopped.
2026-08-27 18:46:00
1
user8576 :
just quantum it
2026-08-31 17:26:13
1
SaoxDxD :
i stiml beleave we can reach a better math equation than transormers or ssm that get better compute or at least more routines in auther way its organizing and needing less max ram and ssd but more compute
2026-09-05 01:11:18
0
woilaehr&d.gguf :
get rid of transformers?
2026-09-03 12:40:22
0
Phillip M :
its just basic power law relationships. its when variance and resolution limits meet diminishing returns, aka compute-optimal frontier.... as said
2026-08-28 21:40:46
3
mmonfred :
Veritasium
2026-09-11 02:29:23
0
.idee.perdue :
it is empirical, right?
2026-08-29 19:26:13
0
Syed Ahmad :
was true a year ago not today.
2026-09-01 21:23:55
0
CheloVerde :
it isn't AI. stop regurgitating their marketing. these are LLMs, they e existed since the 1960s
2026-09-01 14:48:42
0
Rhaokja2k21 :
😳😳😳
2026-09-03 03:58:09
0
To see more videos from user @generallknowledgee, please go to the Tikwm
homepage.
