This but unironically.
Those comments are tight, but really the problem with trying to explain any of this to laypeople isn't exposing how wrong it is. The hard part is making any sense of it.
Like if I told you Donald Trump has connections with a cult that believes grandmothers are a species of raspberry, whose goal is turning Denmark into cheese and oh, a splinter group of theirs just murdered a police officer. That will just raise more questions than it answers. How the hell did they come to believe that? Why would anyone want that? And then I have to choose between looking like a loony conspiracy theorist, doing an impromptu lecture or just having to decide you actually probably don't want to know.
Half of the Finnish Defence Forces aptitude test is just RPM. The other half is a bizarre and inscrutable questionnaire, which includes questions like "is your father a good man" and infamously "would you like to be a florist".
Sounds like a straight (derogatory) version of homoiconicity.
I saw @gerikson's comments after seeing this one and had to click on the tiny thumbnail of his profile pic to make sure it wasn't this exact Squidward face.
If I post this kind of inane drivel, I'll at least make sure the title does not suggest it to be my best work.
"Confiscatory taxes" as opposed to what? Voluntary ones? Just say you don't like taxes as if we couldn't guess already.
To be fair allosauruses went extinct 80 million years earlier so who's really oppressed?
It's only targeting if the card or errata says so.
Shouldn't paint brick walls. Traps moisture.
From what I've previously come to understand, Grimes genuinely does believe LLMs are superhumanly intelligent and rather sinister precursors to an apocalyptic AGI.
Despite this, the output of this GPT is apparently not ominous enough so she needs to write spooky edgelord shit in guise of the bot for attention.
I'm sure she would find some way to cope with the cognitive dissonance if this seeming contradiction was pointed out to her.
bitofhope
joined 2 years ago
It's just depressing. I don't even think Yudkoswsky is being cynical here, but expressing genuine and partially justified anger, while also being very wrong and filtering the event through his personal brainrot. This would be a reasonable statement to make if I believed in just one or two of the implausible things he believes in.
He's absolutely wrong in thinking the LLM "knew enough about humans" to know anything at all. His "alignment" angle is also a really bad way of talking about the harm that language model chatbot tech is capable of doing, though he's correct in saying the ethics of language models aren't a self-solving issue, even though he expresses it in critihype-laden terms.
Not that I like "handing it" to Eliezer Yudkowsky, but he's correct to be upset about a guy dying because of an unhealthy LLM obsession. Rhetorically, this isn't that far from this forum's reaction to children committing suicide because of Character.AI, just that most people on awful.systems have a more realistic conception of the capabilities and limitations of AI technology.