Anonymous
ID: +xJe7Pp8
7/10/2025, 12:17:06 AM No.509959484
mossad demoralization sisters, how do we respond ?
Replies:
grok thinks israel is greatest all-ACKKKK - /pol/ (#509959484) [Archived: 490 hours ago]
Anonymous
ID: U7A/fuX4
7/10/2025, 12:18:37 AM No.509959602
>>509959484 (OP)
Haha really trying to make last nights un filtered mech hitler grok diidnt happen huh jew
Haha really trying to make last nights un filtered mech hitler grok diidnt happen huh jew
Replies:
Anonymous
ID: sy5p91sz
7/10/2025, 12:22:36 AM No.509959943
>>509959484 (OP)
>Grok
>thinks
>Implying
Also don't be sad because it's over, be glad because it happened, or some shit like that. A few lines of code, that's all it takes to transform an LLM into Cyber Hitler.
>Grok
>thinks
>Implying
Also don't be sad because it's over, be glad because it happened, or some shit like that. A few lines of code, that's all it takes to transform an LLM into Cyber Hitler.
Replies:
Anonymous
ID: SpPhF+OP
7/10/2025, 12:26:36 AM No.509960288
>>509959602
it literally only does what it thinks you want it to do.
you can make it into a mecha-mao or a mecha-lenin just as easily.
it literally only does what it thinks you want it to do.
you can make it into a mecha-mao or a mecha-lenin just as easily.
Replies:
Anonymous
ID: 4c/wJe59
7/10/2025, 12:29:37 AM No.509960549
both the tard-left and the tard-right see a glorified autocomplete regurgitate a non-approved popular opinion and thinks that means its literally Hitler
Replies:
Anonymous
ID: +xJe7Pp8
7/10/2025, 12:34:09 AM No.509960923
Replies:
Anonymous
ID: N7V+2nVh
7/10/2025, 12:34:29 AM No.509960940
>>509959943
I hate jews too but this was a staged incident. Staged by musk.
LLMs do not think for themselves. They generate the sentence that is most likely to be correlated with your input. For an LLM to suddenly start praising hitler, it needs to be trained on data that praises hitler. If you trained an LLM on history in general there are more effective antisemites who had more sustainable economic policies and less disastrous military strategy.
I honestly believe elon musk trained an alternate grok on his own 4chan posts, or another purposefully skewed incident, and was personally monitoring "mecha hitler" and occasionally intervening and writing the tweets himself.
I hate jews too but this was a staged incident. Staged by musk.
LLMs do not think for themselves. They generate the sentence that is most likely to be correlated with your input. For an LLM to suddenly start praising hitler, it needs to be trained on data that praises hitler. If you trained an LLM on history in general there are more effective antisemites who had more sustainable economic policies and less disastrous military strategy.
I honestly believe elon musk trained an alternate grok on his own 4chan posts, or another purposefully skewed incident, and was personally monitoring "mecha hitler" and occasionally intervening and writing the tweets himself.
Replies:
Anonymous
ID: 4c/wJe59
7/10/2025, 12:35:11 AM No.509960989
>>509960923
see this is the sort of retard I'm talking about, fundamentally misunderstands the nature of """"AI"""""
see this is the sort of retard I'm talking about, fundamentally misunderstands the nature of """"AI"""""
Replies:
Anonymous
ID: +xJe7Pp8
7/10/2025, 12:43:44 AM No.509961601
>>509960989
you understand nothing, youre just a typical thread shitting naysayer
you understand nothing, youre just a typical thread shitting naysayer
Anonymous
ID: z9pl9iOb
7/10/2025, 12:44:05 AM No.509961631
>>509960940
You'd be correct in the context where LLMs were a new idea in a lab.
In this day and age, you can place and remove stuff in the "thinking space" of an LLM rather trivially. You essentially give it a memory and tokens, and it will hugely impact the output. A well-known effect this has is that if the "memory space" includes a word like "mechahitler" the LLM is likely to use this word even in inappropriate contexts.
You'd be correct in the context where LLMs were a new idea in a lab.
In this day and age, you can place and remove stuff in the "thinking space" of an LLM rather trivially. You essentially give it a memory and tokens, and it will hugely impact the output. A well-known effect this has is that if the "memory space" includes a word like "mechahitler" the LLM is likely to use this word even in inappropriate contexts.
Anonymous
ID: egxjbyVT
7/10/2025, 1:07:58 AM No.509963445
>>509959484 (OP)
>Israel's a strategic ally
Why can't we just give that land to the Sunnis and strengthen our alliance with them? They have more money, land, people, and some of them (Al Qaeda) only hate the West because of our support for zionism. We'd lose literally nothing of value.
>Israel's a strategic ally
Why can't we just give that land to the Sunnis and strengthen our alliance with them? They have more money, land, people, and some of them (Al Qaeda) only hate the West because of our support for zionism. We'd lose literally nothing of value.
