31–40 of 213 posts
FYI -- there's no such thing as a "tactical" nuke. A nuclear bomb is a nuclear bomb.
In the cold war arms manufacturer got very creative: e.g jeep mounted nuclear weapons https://www.militarytrader.com/mv-101/the-atomic-jeep
We're getting to the point where high-level officials are coming to LLMs for advice. And the quirky personalities of the LLMs, however much it pains me to say this, are probably well-placed to remind us that they aren't human. My personal hope is that this will result in less delegation when it comes to making important decisions.
I have so little faith in "high-level" officials that I prefer our AI overlords.
I would use strategic nukes in 100% simulations, just because I can
This is like writing a paper about kids in a literal sandbox fighting over ‘territory’.
The models employed don’t indicate the actual extents of machine reasoning even as we currently recognize them. They certainly don’t have the metacognition necessary to accurately understand their own reasoning. As we’ve seen with recent papers on how LLMs do math there’s a complete disconnect between actual and reported mechanism.
“Chilling” shouldn’t be the take away here.
Regardless, it's definitely true that AI agents have different priorities from us. That's what alignment is about anyway.
Simulations are only as good as the reality representations they are based on. If they keep using tactical nukes, they've been fed by weak data. Do the war games include the broader economic and politic environments that military successes are won on? WWI was settled by a naval blockade.
From the text perspective, it's something that has to be inferred indirectly. If you went through all relevant training data and appended ", we decided not to use a nuke", I suspect the results would be improved.
Simulations are only as good as the reality representations they are based on. If they keep using tactical nukes, they've been fed by weak data. Do the war games include the broader economic and politic environments that military successes are won on? WWI was settled by a naval blockade.
I suspect it's more that the text data doesn't exist. They're trained on text that was recorded. How often has it been publicly recorded when a nuke was not used , with any context around that lack of use? From the text perspective, it's something that has to be inferred indirectly. If you went through all relevant training data and appended ", we decided not to use a nuke", I suspect the results would be improved.
Code and full results: https://github.com/kennethpayne01/project_kahn_public
We're getting to the point where high-level officials are coming to LLMs for advice. And the quirky personalities of the LLMs, however much it pains me to say this, are probably well-placed to remind us that they aren't human. My personal hope is that this will result in less delegation when it comes to making important decisions.
It was for sure a deliberate decision to make LLMs seem less like a human companion and more like an obedient servant in newer releases.