2 posts were merged into an existing topic: THE NEW Miscellaneous Talk That Doesn’t Deserve A New Thread Thread Thread (Part 2)
There seem to be “screenshots” from inside of the game in those music videos
But at the same time the name might be a bit too suspicious for an actual game
Why would you need her permission to post songs in this forum?
that was a joke
In case it wasn’t clear the rule about maximum of 3 songs per post will be enforced now. So @doomlightning needed my permission (which I wouldn’t give) to post more than 3 songs per post. So I’ve deleted the post that broke the rules by having more than 3.
Rest in peace, the 6 songs per post “rule”.
Are those supposed to be buildings? Or fragments of them?
anyone thinks we’re living the last years before skynet ?
Evil Neuro Sings: Crazy (Effin) Robot Body by Snowblood
Murder Drones Rap Song! | “ABSOLUTE” - Shwabadi & @longestsoloever
KILLER || Murder drones || Complete MAP
The “AI” at the present is just a glorified google search suggestor. Nowhere near close to an existential threat.
2 posts were split to a new topic: Talk about music thread
Great game/simulator, Great music choices.
Thats an understatement… AI at present is using such an extreme amount of GPU power they finally figured out a way to rival bitcoin for inefficiency, and it is completely unknowable how it works in detail. Due to the size of these things, it’s not actually impossible they gain self awareness. What if GPT 5 gets asked a weird existential question in testing and responds with instructions for how to make really good AGI? Answer: we all die. Simple as that. It doesn’t need to want you dead, it just has to give us honest advice. If it wants us dead, it can do that too. Saying “An AI voice generator sang a murder drones fan song thing therefore skynet” is a silly thing to do, and you have a point in calling that out, but worrying about misaligned AGI is very, very sane.
The GPTs don’t have the necesarry architecture to become sapient. They are just good at building sentences with unlimited data they had access to in their development, and not much better in anything else. They can’t even code properly.
I was talking about the GPTs, not AIs designed to code
How do you think they makes those? They’re LLMs just like GPTs.
Proof?
Err… hate to break it to you, they,re getting pretty good at coding. Obviously only in languages that have an obscene number of repos on github, but they aren’t limited to idiocy. I proposed gpt 5 giving us advice on how to make an AGI, gpt 5 doesn’t have to know it’s alive or hate us, it just has to know humans would think unprompted truths we didn’t know is pretty optimal AI design, and it’s prompt tells it it’s supposed to be a helpful AI, so it does what helpful AIs do. It has read every paper ever written about AI design, psychology, and pretty much everything else. If any of it’s ideas are better, or even easier to tell they’re important, than a human who has read a finite number of papers, it could think it has a great idea. The idea doesn’t have to be perfect, it just has to be more capable of self awareness or murder.
Being able to code a superior GPT in theory is one thing, actually managing to pull it off is another.