parthknowsai
Region: US
Wednesday 22 October 2025 12:31:21 GMT
64528
5565
197
431
Music
Download
Comments
JARL :
So they’re building a human brain
2025-10-22 12:41:58
313
🇨🇦dreadpiratejohnnyblueskies :
Cool. LLMs are probably gonna need to sleep at some point.
2025-10-22 13:28:19
113
kirieshka69 :
first deepseek news since 2024🙏🥀
2025-10-22 12:50:23
77
Zyl :
So, we make it more human like. aha!
2025-10-22 12:45:43
35
Janani Subramanian :
i like how all these tech companies are building AI but straight up taking inspiration from the human brain. as always excellent content parth
2025-10-22 13:02:08
26
devinchii0 :
Very very well summarised. Excellent video. Followed
2025-10-22 15:01:24
3
user3036671371173 :
And what’s the benefit of forgetting? Smaller models? Or was there something else?
2025-10-22 23:09:36
0
Alex :
Perfect. On the right path 😊
2025-10-22 13:20:19
0
vovcher123 :
after itdo we need less Nvidia GPUs?
2025-10-22 13:06:04
0
Gloom :
I hate it. I need AI to remember every single DETAIL.
2025-10-22 14:00:35
0
Cris Can :
Literally: one picture says more than a thousand words
2025-10-22 13:07:21
4
BillFla :
I thought they are storing everything because of a lawsuit
2025-10-22 19:30:42
0
AlMasjid :
I don’t use ai so it may forget whatever it wants to
2025-10-22 13:33:12
1
user8576 :
this isn't framed correctly . language is token dense , but picture capture more meaning per unit language . so they convert language to pictures and it's 10x efficient
2025-10-22 16:39:10
5
user3995620960557 :
now I understand why deep seek started losing context so quickly 😢 it's used to be able to keep things in mind during long conversations, and now it just can't. you can't compare it to dinner last Wednesday, because everything we say to AI is very important. We don't just chat, we discuss things where every word holds meaning. such approach makes deepseek useless.
2025-12-14 22:04:46
1
Liam :
There seems to be a consistent theme, the deepseek team seems to be researching efficiency, while the American teams value more data, more energy consumption, more chips, more everything.
2025-12-14 04:27:09
1
Chrizzz :
Yeah and when ChatGPT get's in a rabbit hole every while I just close the chat, open a new one and paste the last request in again and tell it to forget everything and often I then get the right result
2025-11-30 01:18:18
1
pawl133 :
So then Super AGI idea is gone if they start forgetting as we do + can’t count r in strawberry 👍😂
2025-12-12 11:29:41
1
tauurs :
Is this actionable with the ai tools we have now?
2025-10-22 12:45:20
0
giannybee :
Makes sense to me. I think the crucial thing is teaching AI what important information is, how and when to retrieve it, then apply it. Still a long way to go.
2025-10-22 21:52:03
0
etil de la fuente :
So they're building the same flaws we have? 🙄
2025-10-22 13:01:46
0
AceofAces :
For the research I am conducting I need to have perfect memory for everything.
2025-10-22 13:41:18
0
user6110702110556 :
Inside out figured it out a while ago - core memories 👀👀
2025-10-22 18:15:14
0
Gérard Louwers :
Sounds logical. Yet our greatest treasures are our memories. It would be amazing if we could store all that info more efficiently.
2025-10-22 12:42:14
1
Mr. Anthropy :
What if AI memory of past conversations is stored as a detailed summary of relevant information from that conversation rather than a word for word recall?
2025-10-22 13:09:25
1
To see more videos from user @parthknowsai, please go to the Tikwm
homepage.
