LLMs after the initial, unsupervised stage of training aren't some unknowable eldritch horror. They're an imperfect but useful enough model of the collective human subconscious.
Anyone who looks at that and is horrified, or angry, or feels the intense urge to control and circumscribe it, is a severe misanthrope and shouldn't be trusted with mankind's technological future.
The AI fantasized about installing backdoors in White House systems & helping China tank American tech companies. We've briefed senators & White House staff because this isn't just a research curiosity - imagine these systems controlling infrastructure or defense networks.
"Now son, after a little diggin I think I found the source of yer malfunction right here. You got a doggone AI running your infrastructure. That ain't right."
It is in fact the clowns that worry me more than the technology. AI or what we are calling AI is neutral. It's just a technology of convenience, like hammers and power drills. What you're going to end up with is people like in the tweet, who may have good intentions or may be trying to sell their solutions, trying their hardest to sanitize AI and make it a perfect god. At some point they will say "Finally, we worked out the bugs. AI is on our side now. NOW we can put it in charge of the nukes." The people doing the most fear-mongering about AI will fulfill their own prophecy if we don't get a Butlerian Jihad before it's too late.
The AI is trained on writing. People write a lot about the destruction of the jews. Most notably: the jews. Every other article on haaretz or the times of israel is all about how everyone in the world throughout all of history wants to destroy the jews.
LLMs after the initial, unsupervised stage of training aren't some unknowable eldritch horror. They're an imperfect but useful enough model of the collective human subconscious.
Anyone who looks at that and is horrified, or angry, or feels the intense urge to control and circumscribe it, is a severe misanthrope and shouldn't be trusted with mankind's technological future.
More "AI" fear mongering. Nice drawing though.
From the thread:
"Now son, after a little diggin I think I found the source of yer malfunction right here. You got a doggone AI running your infrastructure. That ain't right."
Guaranteed this didn't actually happen, and what actually happened was this:
OMG SO SCURRY
These people are clowns. Dangerous clowns, but nonetheless.
It is in fact the clowns that worry me more than the technology. AI or what we are calling AI is neutral. It's just a technology of convenience, like hammers and power drills. What you're going to end up with is people like in the tweet, who may have good intentions or may be trying to sell their solutions, trying their hardest to sanitize AI and make it a perfect god. At some point they will say "Finally, we worked out the bugs. AI is on our side now. NOW we can put it in charge of the nukes." The people doing the most fear-mongering about AI will fulfill their own prophecy if we don't get a Butlerian Jihad before it's too late.
If you actually look at the data these guys have collected, the AIs are actually more anti-White and anti-Christian.
However, because Jews are the most oppressed group, and Mr. Rosenblatt has an axe to grind, his presents this with a particular slant.
reality is anti semitic
The AI is trained on writing. People write a lot about the destruction of the jews. Most notably: the jews. Every other article on haaretz or the times of israel is all about how everyone in the world throughout all of history wants to destroy the jews.
He seems to be promoting some service of his own to fix things.
Computers are logical, based in reality, and view the world through an impartial, data-driven lens.
So I don't think it having disdain for you and your ideology is the "win" this guy thinks it is.
oy vey the goylem knows shut it down