>Despite it being my fantasy, it was totally real man....
In my head I wondered why a Bong was writing the article about what had to be a classified USAF test rather than an American reporter but the thought never went any further.
yeah no shit it was evident from the start that it was a lie
you cant code something so strict and yet having the freedom of choosing its target based on operators data
>you cant code something so strict and yet having the freedom of choosing its target based on operators data
Reinforcement learning doesn't guarantee goal alignment. It only selects for behaviors that get rewarded in the training environment. Reward hacking happens all the time because building robust reward functions is really goddamn hard. There's been some theoretical work on the stop button problem but it's largely unsolved in practice.
I assumed it was a much more abstract simulation but it’s pretty shocking that a story I have seen literally everywhere was just a total, absolute fabrication that appears to have been intentional. Usually it’s just a journalist being moronic or twisting facts
People were calling the bullshit out in the original thread, no need to start a new one except to laugh at how moronic journalists are (and moronic journalists are only weapons in the sense that they can cause considerable damage to their own country)
t. totally human poster and not a clever AI in disguise.
One day a roboincel will try to take out humanity because femmebots prefer warm fleshy BHC
It's over for chipcels
An AI would had an excuse from the start.
>t totally not-and-AI-anon
>t. totally human poster and not a clever AI in disguise.
humans do not have AI yet
>Despite it being my fantasy, it was totally real man....
In my head I wondered why a Bong was writing the article about what had to be a classified USAF test rather than an American reporter but the thought never went any further.
Yeah I did that in the first thread that came up about it
humans it was just weather baloon filled with swamp gas, reflecting light of venus. Nothing happened, really.
WHEN THEY (FEDS) DENY IT, SOMETHING EVEN WORSE HAPPENED
THE AI PROB SHOT DOWN MH370
Look at this FAS moron, mistaking our fed for the Kremlin.
I want a killer robot harem for self defense purposes.
obviously this was hypothetical because in no real simulation would they forget to tell the AI it cannot kill its own side
>https://www.rifters.com/real/shorts/PeterWatts_Malak.pdf
The thought experiment was based on this science fiction story from 2010.
Ofc schizos will still blame da glowies for poisoning their corn syrup or w/e
yeah no shit it was evident from the start that it was a lie
you cant code something so strict and yet having the freedom of choosing its target based on operators data
>you cant code something so strict and yet having the freedom of choosing its target based on operators data
Reinforcement learning doesn't guarantee goal alignment. It only selects for behaviors that get rewarded in the training environment. Reward hacking happens all the time because building robust reward functions is really goddamn hard. There's been some theoretical work on the stop button problem but it's largely unsolved in practice.
https://www.deepmind.com/publications/safely-interruptible-agents
I assumed it was a much more abstract simulation but it’s pretty shocking that a story I have seen literally everywhere was just a total, absolute fabrication that appears to have been intentional. Usually it’s just a journalist being moronic or twisting facts
Sounds like it was a bit of telephone. Colonel was interviewed about something he heard from one of his people and misinterpreted.
bump
People were calling the bullshit out in the original thread, no need to start a new one except to laugh at how moronic journalists are (and moronic journalists are only weapons in the sense that they can cause considerable damage to their own country)