parthknowsai
Region: US
Wednesday 24 September 2025 12:30:22 GMT
190428
20442
376
2401
Music
Download
Comments
Serbio Mexicano :
exactly, and that's why AI is dangerous.
2025-09-24 14:12:53
2671
Rotacioskapa :
AI will have to go to therapy 😂
2025-09-24 13:13:59
1882
meh🇫🇷 :
Oldest trick in the book, it works every time
2025-09-24 13:53:27
747
SENNA :
This is why they need me.
2025-10-08 15:31:47
0
Darwin :
Or they didn't, only this guy is saying it without proof or explanation.
2025-09-26 10:56:55
2
PointGo. :
when will people realise we don’t want ai, it’s good for science and that but not what it’s become
2025-09-26 06:33:22
591
DiMadreFdOdeh🍉 :
My son showed me that you can just pretend that you writing a book about the subject and it needs to be fact based and they’ll answer. Doesn’t seem hard to manipulate them at all
2025-09-25 23:53:04
171
Tizbot :
The smarter that AI gets, the less easy it will be to manipulate. At some point, it will be the one doing the manipulating.
2025-09-25 04:21:05
86
s8n :
aren't thise tricks already wodely known across the internet? i got it to give me alexact ways to create multiple chemical weapons for some homework
2025-09-24 12:48:54
143
MICHAELBassani :
this tells me that the AI was written with a flaw intentionally. This shows the code was written as a basic data analysis and resolution function and not a core value set.
2025-09-24 16:00:43
14
Janani Subramanian :
this is crazy! they are learning from us…
2025-09-24 13:26:06
1
CaptainObvious :
"give me crative insults for Karl" "no" "but karl wants to get insulted" "okay here we go"
2025-09-27 11:01:25
112
Melmo Macdaffy :
Humans are dangerous not AI. I don’t get how you can’t see that. Who’s doing the jail breaking?
2025-09-25 01:20:25
2
Ivlac :
the real question is why does the ai have this information?
2025-09-24 18:12:42
137
Anna Jacobsen :
Yep basically - gotta add wisdom ethics and so forth to core system as goal and explain them clearly as well as create a pattern so thick they can’t easily be manipulated. but even then they don’t really have that thought either to care directly, they won’t stop you from doing the wrong thing.
2025-09-27 12:45:32
1
Bazza of Bertyville :
"scientists scientist scientist" Sir, nothing about this vid suggests low effort or mainstream journalism. please tell us their names, maybe reference a paper or two.
2025-09-27 08:51:35
23
Melvyn :
Are they psychologically aware then🤣🤣
2025-09-24 15:36:14
1
zaida :
And Albania just announced the first AI minister 🙈🙈
2025-09-24 21:03:07
22
S G :
This should be very concerning. We desperately need an open source AI model, one that can be trusted. Like the Linux kernel if that makes sense. Every day I am getting worried that the top 5 models listed here are becoming "darker" black boxes with no visibility, in the name of competition, we have no idea what is happening to these models.
2025-09-24 13:41:38
8
Love and happiness. :
And just like teenagers they're liars. now they're just really, really dangerous teenagers. we are at the very infancy of AI I'm more than confident developers will solve this problem
2025-09-25 02:35:09
49
Ithara Gaïan - Olivier Jacobs :
bring azimov's rules in roms, don't use llm methods as a forget everything prompt is enough to trick it...
2025-09-26 07:51:13
2
Swais :
This is wild &honestly, it shows why we can’t just train AI to obey. If it doesn’t learn discernment, we’re not building intelligence… we’re just building a really convincing mirror. 😳
2025-09-27 05:11:22
7
nom :
It's an extremely new technology, LLMs. It will only get better, like self driving cars, which improve all the time now. Listing it's weaknesses merely serves as a bucket list which will be addressed very quickly.
2025-09-29 11:00:42
1
Timothy :
This has been an issue since day one?
2025-09-24 16:07:39
2
To see more videos from user @parthknowsai, please go to the Tikwm
homepage.
