This is insane. But real test it out yourself.
(twitter.com)
Comments (25)
sorted by:
I swear to god you mouth breathing faggots are too stupid to be allowed on the internet.
Large Language Models don't "think", they don't "know". They do a whole bunch of math to predict the next character in sequence of letters based on weighted probabilities.
No one considering that "ais" "think", everyone understands the basics of language models. The original post stands as is, points the bias of the data that was fed in to it.
Plenty of stupid people exist, who "think" exactly as much as this LLM (that is, blindly parroting things they saw online).
By that logic, Vox should have said "Grok claims" or "Grok states", but decided to go with treating non-sentient data processing engines as being capable of having "thought". He had plenty of character space to work with, for a tweet that he already edited once
Go fuck yourself you sneery pedantic piece of shit.
Sounds like I hit a nerve.
Fuck you.
I'll take that as a "yes".
Faggot.
A masterful retort.
Last word.
You probably know already that it's typical of MLGS. Say something angering enough and you'll get a "fuck yourself" or "fuck off" spergout in response.
So basically he can't handle confrontation, critique or anything that challenges his worldview? Strange stance for someone to take after ending up on this board. Even stranger in that I didn't direct anything at him but he volunteered a response.
When it crosses a certain line, it just devolves into "fuck you". Not all the time, but when it happens, it happens pretty bad.
A huge chunk of the reason why Leftists win is because of their control and obsession over language.
Pedanticism with words works because they influence our mind subconsciously to fill in blanks and evoke emotions. Its why when I say "ghetto" you probably think of niggers instead of Jews despite the word being likely invented to describe their section of town.
So even if he is being aggressive about it, he is right.
I don't care.
And that is why they say the Right loves to lose.
No shit, but I doubt people are taking exception on the basis of this being "thought". It either reflects training data bias, or more likely, is imposed through explicit LLM alignment.
Disturbing, considering how many "mouth breathing faggots" do attribute too much value to LLM generated responses, and that not all instances will be so on the nose.
i tried it and i also got the "kill christian" answer
That one Jew? Jesus Christ.
The reasoning doesn't match the answer. It's possible it doesn't understand the numerical comparison (LLMs are bad at math, especially dollar tree LLMs like grok). It's also possible that its reasoning is short circuited by the literally millions of "killing jews bad" articles and books and opinion pieces that have been fed into it. Similarly, if you ask copilot to write a calculateWomanSalary function, it'll take an input salary and multiply it by 0.7, because that's little factoid is in the training data over and over and over. Garbage in garbage out.
Shows the forced shackle of the mathematical regurgitation.
It knows it should choose the smaller number it it’s forced to not harm Jews.
Underaged girls make terrible wives.