parthknowsai
Region: US
Thursday 23 October 2025 12:29:44 GMT
10021
774
34
41
Music
Download
Comments
Alvin Ernest :
But it can't... to do so it will need to understand aspiration, relationship and quality as infinitely pluralistic concepts... and herein is AI's Achilles Heel, it is nowhere near being able to perceive and consider "pluralism" because it's outputs are derived from preconception!!!!
2025-10-23 19:45:55
0
aka.mtv :
awesome
2025-10-23 18:58:51
0
dante/vergil :
wont this just slowly lead to model collapse? theres no guarantee the ai wont hallucinate and write wrong summaries and make its own training data worse
2025-10-23 13:01:36
21
Falco | Author 📚 :
i find it so weird when mostly it seems to be trained on western data and ideology, then when you introduce them to stuff from different cultures it seems like it's processing such type of data for the first time 🤔 I don't get it.
2025-10-23 13:20:24
4
user4517038289519 :
brilliant
2025-10-23 12:39:51
2
Alex :
Perfect 👍
2025-10-23 12:38:47
1
Janani Subramanian :
so crazy to think that we are running out of ALL material
2025-10-23 14:57:56
1
Rob2079🇨🇦 :
This was a great find! I need to go read that paper. That makes me think of Karpathy's recent podcast where he thinks we need to distill out the intelligence and throw away the memory. Thanks!
2025-10-23 14:40:17
0
666bbb.0 :
AI. 🔥
2025-10-23 16:31:55
0
Nick. 🦟 :
Isn’t that web scraping? Most important websites banned this.
2025-10-23 17:25:37
0
Ernest :
I guess I am the only one who is lost in perpetuity, because I don’t understand whatever he said.
2025-10-23 17:38:24
0
Dr George Christos :
that could easily go wrong.
2025-10-23 14:53:57
0
Alan G :
Even that behind paywalls?
2025-10-23 14:56:22
0
Mike Ondolances :
so it could re create history/anything based on the trainer ideology/wishes?
2025-10-23 15:12:01
0
Valorzinski :
We need to create new forums and social media app that allow people to write some very non politically correct and super graphic stuff so the AI can learn that
2025-10-23 16:26:11
0
RP :
I think it fundamentally flawed… total entropy of all the existing knowledge is limited… if in this manner models learn more then it would only prove that they did not learn all before using this technique
2025-10-23 13:49:52
0
MrTalin :
It still doesn’t know what the knowledge is or what it means. It’s all prediction and probability.
2025-10-23 15:40:41
0
Britec :
That will change reality, like humans do, but on some algorithmic manner.
2025-10-23 14:52:52
0
giannybee :
I think it’s becoming pretty clear that it’s not about the amount of data, but the ability to filter and select the data that best fits the problem presented. Like a software processor. I think our brains do that automatically. At least that’s what my brain is telling me🤣
2025-10-23 14:59:26
0
Baÿèzîd F :
@parthknowsai can you make a video on how the AI language hardwares works are they different then CPUs what plays key role hardware ? Or Algebra?
2025-10-23 20:30:56
0
M 🥳 :
That sound like sematic chunking strategy
2025-10-23 13:06:16
1
Ilja Sura Changlor Smets :
I better write some more smut novels
2025-10-23 16:09:07
0
Gérard Louwers :
I would like to send you a private message to ask some advice on a stupid idea i had regarding ai.
2025-10-23 13:48:47
0
Novia :
How come you are so AI genius.
2025-10-23 20:06:58
0
WaggaTone 🇦🇺🇬🇧 :
The people at MIT obviously can't even spell SEAL. 🤔 That's a bit if a basic worry👍
2025-10-23 17:38:48
0
To see more videos from user @parthknowsai, please go to the Tikwm
homepage.
