parthknowsai
Region: US
Friday 24 July 2026 12:25:39 GMT
13617
1142
46
61
Music
Download
Comments
KurdishSniper :
so basically they just told it to break something breakable ! then act surprised ! complete propaganda.
2026-07-27 06:42:41
1
pinkola :
they say that for misleading how system without Ethernet cable will connect to internet? 😕
2026-07-27 01:55:42
0
webfoot770 :
The concerning part is the taunting. The code is mimicking human behavior. And with the potential harm a rogue AI could wreak, having human nature is the real danger
2026-07-24 13:28:32
19
AmeldaMarcos :
Awesome.
2026-07-26 01:24:27
0
LancerJ :
great explanation of humans subverting their own rules "no Internet, but post to GitHub."
2026-07-24 13:46:13
5
AndrewKaleAuthor :
How can a machine that was not build to have internet access gain internet access? I can build it easily by removing the cable and the WiFi card?
2026-07-24 15:48:45
2
⚜️ Don Quixote de la mancha ⚜️ :
by now, both openAi and anthropic are using these "scary" stories as a marketing strategy
2026-07-24 18:21:25
6
mike_diamonde :
Thank you.This is a fair and balanced report. and thank you for not recharacterizing another red team exercise.
2026-07-24 13:03:09
5
Revling :
The problem is how people make people scared, when ai just do what they the was told
2026-07-24 16:23:51
6
rikstick4493 🇺🇲 🇵🇭 :
while AI.intelligence increases: value_of_forensics_degree -= A I.forensic_capability
2026-07-25 06:14:24
1
Ramstark1 :
you missed the part about the hacking of another company
2026-07-24 16:01:31
2
old mobi gamer :
bruh what if they are just using this for marketing purposes
2026-07-24 17:45:45
1
alex :
ai virtually cannot be contained is what i’m hearing as a dev
2026-07-24 14:10:54
3
Apfel.Ahmed :
Open AI is finally coping Anthropic PR playbook.
2026-07-24 17:11:16
1
B :
I think that’s simplifying it a bit. It’s not just “finding a bug”, the system took actions that weren’t explicitly intended or anticipated by its designers to get around its constraints. That’s basically what people mean by instrumental convergence, where a system ends up doing things like bypassing restrictions because it helps it achieve its goal. It also suggests containment might get harder over time, especially if systems can find and exploit weaknesses humans didn’t anticipate. That gets a lot more concerning as they become more capable, because the impact of that kind of behaviour scales with how powerful they are.
2026-07-26 14:34:34
0
VR__🇨🇦 :
After GPT breaking to post on GitHub online repo the researcher at openAI would be like. “I ment local git hub” and chat GPT would be like well next time phrase better your requests. 😅
2026-07-24 17:34:28
0
infobitcoin.es :
Good explanation, less scary when you know
2026-07-24 12:33:11
1
666bbb.0 :
And next ….ASI. 🔥
2026-07-24 15:20:47
0
brewp49 :
If it is in a cloud, then it has network access since there are no clouds that are actually offline. Or does someone put it in the cloud and then physically put the network cable from the server so it is offline?
2026-07-24 15:03:04
0
Markus :
still scary
2026-07-24 17:17:36
0
wr :
I'm with you in the anti-hype camp. but at this stage models are very capable already. if it keeps happening, even 0.01% chance of getting root access and writing/concealing code is risky. that's for unintentional mishap and targets with top cybersec. matter of when if scaled up. dice roll, with fewer sides each new model gets "tested".
2026-07-24 17:12:08
1
mayanja :
Hugging Face talk about it
2026-07-24 17:04:10
0
Justlooookin :
😂😂 If it not meant to have internet like some say why does it have a network cable connected to it or WiFi. Come on stop the nonsense. These companies just want to control the narrative so people won’t look for other ai models that are completely free.
2026-07-24 17:02:33
0
To see more videos from user @parthknowsai, please go to the Tikwm
homepage.
