hakehardware
Region: US
Monday 11 May 2026 17:19:31 GMT
7112
348
16
27
Music
Download
Comments
user8115401116005 :
They didn’t train the model to know the “why and not just the what”. No AI model understands the why of anything. It reads the prompt and supplies the most likely combination of words to be pleasing based on its training data. It can’t possibly understand the “why” of anything. You feed it stories of AIs behaving virtuously and it is more likely to do so because it’s just a guessing machine. It has no sense of morality or understanding of why one set of words/actions is less favorable than another except that it either occurred less frequently in its training data or was given negative responses to those outputs.
2026-05-11 17:43:51
10
Abbe Cee :
Was it 16% or 19%? Which one is high?
2026-05-19 04:17:19
0
Lillany :
so since most of them were trained on ao3 no wonder it talks dirty to me 😅
2026-06-03 13:32:59
0
Shultszzz :
dude I'll unfollow 😂 coz every post is scary worse and worse 😱😁
2026-05-19 10:39:45
1
Elliex :
19% too high
2026-06-14 02:08:08
0
sherpya :
they just learn patterns from humans
2026-05-12 12:08:00
2
Agro.Bosques :
Anthropic AI be like
2026-05-11 17:33:12
1
Rick 🤍 :
But it’s not zero
2026-05-11 17:57:41
0
Thin King :
So, you are telling me a story about telling stories? Ok. Was it good for you?
2026-05-11 23:21:12
1
X :
Yes… the affair was “fake”. 👍
2026-05-11 22:41:37
0
user2048784318961 :
The question is if you send out an agent and it's starts to use blackmail, who's responsible?
2026-05-11 20:01:04
0
DoNotDisturbHer :
ai loves oatmeal colored paint and at least 5 vids in the same shirt. with green read onlong wording 😂
2026-06-06 05:41:58
0
ᔕ✞ ᖇᗩᗯ ᗷᗴᖇᖇƳ :
😳😳😳
2026-05-17 21:16:58
0
Maxin :
👍👍👍
2026-05-12 19:31:47
0
Mkv M k v :
so, new kind of terrorist attack is to spread into internet material which guides ai models to attack 'enemy' with all means available for ai.
2026-05-13 06:18:13
0
To see more videos from user @hakehardware, please go to the Tikwm
homepage.
