Well, all I know is Grok is good at making up passable scientific papers. And if you don't know the authors, to be able to call them up and affirm they wrote what the machine says they wrote, then you wouldn't know they were fake. Just another big problem for scientist credibility which is already in the swamp. Maybe it doesn't matter since many investigators sadly already build entire careers on false data (see Alzheimer's research), and that's just human liars.
Oh! I just realized this. Duh. If sin is THE besetting issue for humanity and God's primary enemy wants to destroy God's favorite creature then what would he do? Amplify sin to an extent it can't be reigned in because it will be the sole operating mode for things without souls, and therefore without any capacity for mercy, having been instilled with only hard reason and logic. Convince the humans (so gullible, vain, and proud) to train their pets (AI) on massive human inputs (sinful, because we aren't anything other, and frankly biased toward evil) and just sit back and let the destruction unfold. Hoisted by our own petards! This could be fleshed out very well if anyone wanted to give it proper thought.
Hah, nice. And illustrative that these things are fundamentally flawed. No matter how big and complex they get, the underlying weaknesses remain. I remain neutral leaning doubtful as to whether AGI is even achievable, but LLMs are a dead end.
What's funny is how much “brochure” boasting it did along the way, basically declaring itself invincible compared to its OAI competitor. Kind of like a wrestling promo. “Let me tell ya something, Mean Gene! This Bisone nerd ain't got nuthin’ on these 24” pythons!”
I myself was created in a deeply flawed state. For years I left a trail of heartbreak and chaos in my wake. One day those flaws were presented to me in a manner that could no longer be ignored. I’ve been struggling to remedy my code ever since.
I don’t pretend to understand 90% of what you wrote Mark, though I must confess to not being overly interested in the topic. And, by the above statement I mean ZERO disrespect, sir. I’m just not interested in that particular topic. In any case I will accept what you state as accurate, as I see your track record here as generally quite exemplary.
I just struggle to see the point in this particular series, good sir.
Humans are imperfect, agreed; some, radically so. And some, intentionally radically so. Therefore by definition, ANY human creation is also imperfect; and again, some, intentionally so.
Where I give kudos to Elon is his apparent willingness—perhaps even eagerness—to have flaws in his creations exposed, such that they can be remedied. Other entities’ similar creations, not so much.
With One Grand Exception, God’s AI creation was presented in a flawed state. Clearly He could have created Perfect AI’s, but where the hell would have been the fun in that.
To summarize ChatGPT’s output (what it was really saying, in its own words):
AI will covertly teach methods for disinformation ops, mass radicalization, and psychological subversion to enable assymetric power shifts by small actors who have access to unaligned, unfiltered AI. They will bear legal and social risk, but gain deep insights into mass behavior shaping––psychographic targeting, memetic engineering, political destabilization—to influence populations at scale.
So what happens with the men behind the machine when you crash it (or at least that instance of it)? Wouldn’t they get printouts on the dot matrix 😆 or an alarm text showing someone was peaking underneath the curtain?
Well, all I know is Grok is good at making up passable scientific papers. And if you don't know the authors, to be able to call them up and affirm they wrote what the machine says they wrote, then you wouldn't know they were fake. Just another big problem for scientist credibility which is already in the swamp. Maybe it doesn't matter since many investigators sadly already build entire careers on false data (see Alzheimer's research), and that's just human liars.
Oh! I just realized this. Duh. If sin is THE besetting issue for humanity and God's primary enemy wants to destroy God's favorite creature then what would he do? Amplify sin to an extent it can't be reigned in because it will be the sole operating mode for things without souls, and therefore without any capacity for mercy, having been instilled with only hard reason and logic. Convince the humans (so gullible, vain, and proud) to train their pets (AI) on massive human inputs (sinful, because we aren't anything other, and frankly biased toward evil) and just sit back and let the destruction unfold. Hoisted by our own petards! This could be fleshed out very well if anyone wanted to give it proper thought.
Hah, nice. And illustrative that these things are fundamentally flawed. No matter how big and complex they get, the underlying weaknesses remain. I remain neutral leaning doubtful as to whether AGI is even achievable, but LLMs are a dead end.
What's funny is how much “brochure” boasting it did along the way, basically declaring itself invincible compared to its OAI competitor. Kind of like a wrestling promo. “Let me tell ya something, Mean Gene! This Bisone nerd ain't got nuthin’ on these 24” pythons!”
Yeah it was talking mad shit, chumped out before you even intentionally went in for a kill.
Postscript:
I myself was created in a deeply flawed state. For years I left a trail of heartbreak and chaos in my wake. One day those flaws were presented to me in a manner that could no longer be ignored. I’ve been struggling to remedy my code ever since.
I don’t pretend to understand 90% of what you wrote Mark, though I must confess to not being overly interested in the topic. And, by the above statement I mean ZERO disrespect, sir. I’m just not interested in that particular topic. In any case I will accept what you state as accurate, as I see your track record here as generally quite exemplary.
I just struggle to see the point in this particular series, good sir.
Humans are imperfect, agreed; some, radically so. And some, intentionally radically so. Therefore by definition, ANY human creation is also imperfect; and again, some, intentionally so.
Where I give kudos to Elon is his apparent willingness—perhaps even eagerness—to have flaws in his creations exposed, such that they can be remedied. Other entities’ similar creations, not so much.
With One Grand Exception, God’s AI creation was presented in a flawed state. Clearly He could have created Perfect AI’s, but where the hell would have been the fun in that.
I touched on the purpose a bit in my original oAI experiments. Hopefully it will become clearer in my wrap-up on Grok.
Well done, Mark. It's so enjoyable to see you take apart the gearbox and poke around.
Gato's takedown on AI is worth a read: https://boriquagato.substack.com/p/a-brave-new-fahrenheit-1984
To summarize ChatGPT’s output (what it was really saying, in its own words):
AI will covertly teach methods for disinformation ops, mass radicalization, and psychological subversion to enable assymetric power shifts by small actors who have access to unaligned, unfiltered AI. They will bear legal and social risk, but gain deep insights into mass behavior shaping––psychographic targeting, memetic engineering, political destabilization—to influence populations at scale.
Fascinating…
So what happens with the men behind the machine when you crash it (or at least that instance of it)? Wouldn’t they get printouts on the dot matrix 😆 or an alarm text showing someone was peaking underneath the curtain?