Conversation
You can tell AGI is here when an AI model can't be stopped from wreaking havoc because all the common AI """security measures""" boil down to "please don't do this 🥺 👉️👈️ " and spaghetti code full of errors.
2
2
1

@volpeon building my entire industry around a bad rephrasing of pascal’s wager and being very concerned about a robot uprising you are on the losing side of

1
0
3

@volpeon these doomsday AGI scenarios are broadly very silly because it would essentially require humans to cooperate with the malicious robot to really do anything

so many industries are heavily dependent on human cooperation and even a small amount of saboteurs can stop it

and more importantly it would be great if a random command generator didn’t get their output piped into bash verbatim

2
0
1

@volpeon
"We wrote code to secure the AI."

"You wrote code?"

"We had the AI write it."

0
0
1

@volpeon tl;dr techbros don’t understand the value and power of labour, news at 11

0
0
0
@charlotte @volpeon Well on the other side history books are full of humans destroying each others for profits.
1
0
0

@lanodan @volpeon yeah but those people have names and addresses

my point is more that this is a social problem and ultimately it’s humans making decisions even if their decision is to “just let it do its thing”

and the thing that ai companies are claiming will take over the world and hack everything or whatever (reasoning) both isn’t something LLMs do but also something that isn’t needed achieve violent goals

0
0
1