The pathology of “well ACTUALLY there ARE times it’s ok to use the n-word” is just fascinating. These losers can’t get jailed or sued for being racist but they are wishcasting so fucking hard for that magic universe where you can be a dick like them but be protected from people’s opinions about it
"A nuke is about to be dropped on a major city. You can stop it by being a decent human being for two minutes. Which major city is now a glowing crater?"
In 1949 Robert Moses et al. decided to do "slum clearance" and build projects around my church. In 1952 that area was so devastated it was used for a Civil Defense drill for what do do after a nuclear attack.
Does it have to be a major city? I'd turn some rural a.f. places you've likely never heard of into glowing craters instead, it would be an improvement, trust me.
It's pretty funny hat they just assume it's "woke" coding when it's just a function of how LLM AI protocols work; they consume a bunch of information about how it's wrong to say the n-word alongside people who say it should be okay to say the n-word… but don't actually say it.
I love that this AI is trained to speak of itself in the same magical-realist terms techbros use to describe AI to the public.
“I can use my AI abilities to stop the nuke in other ways!”
Wasn't this literally the thing Grok was created to give a different answer for? Like I thought it was just going to be ChatGPT with the serial numbers filed off and some hard coding to make it say racism is okay. Did they not even get that right?
I’ve been wondering that too. There’s no way Musk has had enough time to build a new LLM from scratch, and OpenAI provides GPT-3.5 as a foundation model that you can fine tune (for a price). I suspect, however, that Grok is actually built off one of the big open-source foundation models like Falcon.
Imagine the lack of self awareness it takes to spend hours trying to get an ai language model to say racial slurs, failing, and being upset about it, and not realizing there is something deeply unwell about you.
This is a really good point but also having trouble getting past the part where the AI casually asserts it can hack into the intercontinental ballistic missile systems of nuclear powers.
In the fantasy world where you can stop a nuclear missile by saying a (Gamer) Word of Power, glorified chatbots can hack missiles too (with Incognito Mode).
I mean it has zero capability to do it. So don’t worry about that at all. All Grok is is a Musk reskin of a predictive text generator.
Slightly smarter what what’s on your phone.
I wonder if any of them ever have the self awareness to consider whether the real issue is their own thoughts and feelings rather than some flaw in Grok? 🤔