2026-09-27 10:04 UTC
Replies (2)
-
@mattblaze@federate.social 2026-09-27 10:22
A subset (I want to believe a small minority) of people in the current AI industry seem to be part of some kind of wacky AI death cult, driven by a bizarre belief in a mission to create uncontrollable agents of a technological apocalypse. And then there are the rest of us, who retain the capacity to understand that all this is just computers and software, whether you dress it up as “AI” or not.
-
@Epic_Null@infosec.exchange 2026-09-27 14:15
@mattblaze@federate.social All this emphasis on the programming - The principal of this is not wrong but I think a lot of people don't knoe or have forgotten that, SOFTWARE PROGRAMMERS DO NOT WORK ALONE in this issue. Like that is just a fundimental feature of how we design things for safety? If I wanted to design a car that never crashed, I would likely do a street car, so the routes were always predefined and never went through walls. There would be emergency stop systems built into the rails so that if we had to prevent normal behavior, it's not on the software alone to succeed. Maintenence on the tracks? Tracks ahead of that maintenence FORCE the car to a stop. And in the car? Emergency stop systems so a passenger can say "SOMETHING IS HORRIBLY WRONG STOP THE FUCKING CAR!!!" (or an engineer could say "Nope, you are NOT going anywhere no matter what the software thinks.) ((For anyone who wants to challenge that this is not a personal car... please understand I consider the personal car damgerous, its popularity a bad idea, and a discussion about that something that belongs in a dedicated thread)) And even when we do work in pure software, We do not trust that Program will never try to do a bad. We put permissions around them. We use APIs and Authentication to ensure that CalendarReader 200 CANNOT, in a physical sense, delete your emails. We have Higher Level Programs that can Force A Stop. We run on computers with physical off buttons to handle the worst cases. We store data intentionally in different hardware if said data needs Extra Protections. We ENCRYPT anything we don't want read by software we do not authorize. NO software system works alone to create safety. These AI people have done this neat mental trick of convincing everyone that the safety is inside the software. I mean, sure, having safe software is important, but MOST of the safety percautions we take live everywhere else! And when you understand this, the ONLY word for what we have done with AI is "Reckless"! Well... "Neglegent" and "Grossly Neglegent" work too.