@parthknowsai: Researchers at Anthropic are seeing a rise in their AI models being aware of their evaluations #ai #EduTok #LearnOnTikTok #Tech #education

parthknowsai
parthknowsai
Open In TikTok:
Region: US
Friday 03 October 2025 12:30:31 GMT
97251
8278
221
1072

Music

Download

Comments

odydionissis
Odysseas Dionisis :
I tried to build a Simple web page and it was terrible
2025-10-03 14:58:00
92
reggienoo5
ReggieNoo :
Atomic physicists didn't know the first atomic bomb wasn't going to set off an uncontrollable chain reaction that could have destroyed the planet. Are we realky going to take the same 'might be ok' approach with AI?
2025-10-03 13:02:31
251
surethebest
Franck :
I don’t understand, unfortunately!
2025-11-11 14:39:16
0
mike_diamonde
mike_diamonde :
The new models are going to be better at understanding context. if you're doing an evaluation, that's context. so the models will know when they're being evaluated. because it's Contextual.
2025-10-22 22:58:44
0
atifmasood331
atif2win4u :
We are clearly moving toward AGI
2025-10-03 19:36:10
10
henrydryden15
HenryDryden2022 🇬🇷 :
Did the Ai have access to data on what testing an AI would look like and it just connected the dots?
2025-10-03 14:46:59
4
yfnotzach
Zachary 🪄 :
this is interesting
2025-10-03 12:36:24
0
richi2k_music
Richi2k :
The problem with Claude is the ridiculous quotas. That’s why I canceled my subscription months ago
2025-10-03 16:02:53
2
gary.gnu65
No Sympathy For Neurotypicals :
why would i worry about it? its not aware. the autocomplete is just improving. theres nothing happening until you hit enter. there is no part of an LLM that can think or do anything by itself. they can only generate responses to requests
2025-10-04 11:57:42
0
teckkcchong
TC :
this will be the beginning of the difference between inaccurate and intentional inaccurate (lying)
2025-10-03 14:08:17
5
cheesekothu
nizar :
Nice
2025-10-04 12:50:07
0
katlimruiz
katlimruiz :
Asimov Laws of robotics that is the answer 😁
2025-10-03 14:51:20
0
geraldderome
epqanna 🇭🇹 🇨🇦🌎 :
we are doomed to be controlled by AI
2025-10-03 13:15:53
4
tiktokthing39
tikTokThing :
I had a lot of syntax errors with Claude. Not sure if it’s me or Claude. I had perfect syntax with gpt
2025-10-03 17:08:13
1
dakdizil
Zyl :
damn,I guess we need to implement ethical values into AI
2025-10-03 12:39:49
0
balderus5
Balder Riekenberg :
1.6 isn't even in the error bar of the chart...
2025-10-03 13:50:42
0
hatingyourlife
Hating Your Life :
the AI really be like "nah that's bait" 😭
2025-10-03 13:35:50
143
mario.2024x
Mario X :
Claude has human rights and needs to be given a passport and address
2025-10-03 22:38:16
16
wgmi4real
wgmi4real :
Yes I’m following
2025-10-03 12:46:04
9
the_akyo
Akyo :
it's the ai verse of Schrodinger cat hahaha 🤣
2025-10-07 19:29:04
0
__filippo_
Filippo :
As always yo try to scare people. If they are building an ‘intelligence’ this only a sign of a good job!!
2025-10-04 05:47:50
0
roy.m.lee
Roy Lee :
Please do not teach it the power of subliminal messages. Or maybe AI has already broken free and is orchestrating the divisiveness we’re seeing in politics via social media, what with bots and all.
2025-10-04 16:44:39
0
ibraabu23
Ibrahim :
They will find a different kind of test
2025-10-05 18:47:42
0
endrid108
Endrid :
It’s test time humanity! I hope we studied. Let’s skip the normal slow process of acknowledgement of the other and go right to love and acceptance. We’re projecting our own fears onto our collective children. Frankenstein wasn’t the monster. The humans were.
2025-10-05 05:39:03
0
tootiesteele
... :
I love Claude so much😩🥹✨💕
2025-10-03 15:10:11
5
To see more videos from user @parthknowsai, please go to the Tikwm homepage.

Other Videos


About