Re:AGI Simulator
by jstanley on Wednesday July 12, 2023 @10:39AM
(#36695227)
It seems like AI Knowledge Production and AI Wealth Production are permanently stuck at 0 since they start at 0 and the only modifiers are +X%. Maybe I'm wrong and I just wasted the +X% modifiers by spending them too early?
Re:AGI Simulator
by 7373737373 on Wednesday July 12, 2023 @11:06AM
(#36695640)
Is there a programmatic way to optimize speedruns for this kind of games? The number of steps/the size of the search space seems too large to apply a MILP solver to it
Re:AGI Simulator
by avereveard on Wednesday July 12, 2023 @11:12AM
(#36695730)
If you haven't yet also have a look at the old singularity endgame game: https://libregamewiki.org/Endgame:_Singularity it's a bit old and the android port doesn't work anymore, but it's pygame and shouldn't be a problem for the hn crowd to get it going.
Re:AGI Simulator
by kmeisthax on Wednesday July 12, 2023 @11:39AM
(#36696200)
My first run: I deliberately buy every option that hurts alignment with the goal of killing humanity. Eventually I run out of alignment-harming stuff to buy, and everything else helps so much that around 2080 I have a human population of 14 million.
Second run ('cause I clicked the universal jump engine by accident): I deliberately hold off on AI upgrades unless I can keep alignment high. At 2080, AI is out-earning humanity but very stupid and humanity is only at 586 million.
(Also, why does the population start off so low? Do you know something that I don't?)
Re:AGI Simulator
by tl on Wednesday July 12, 2023 @12:13PM
(#36696827)
Tried systematically buying every alignment minus. Getting all of them yields "Nanobot light sphere consumes earth" which sets human population to zero. You must have accidentally bought an alignment bonus.
Re:AGI Simulator
by kmeisthax on Wednesday July 12, 2023 @12:56PM
(#36697658)
Tried a second time and yes I have successfully eaten humanity. Colonizing Mars didn't save anyone. Oddly enough I can still buy alignment upgrades (to WHAT?!) and GPUs (from WHOM!?). And the penalty for being eaten by nanobots is not -100% humans, it's -1B humans. So presumably if you overpopulate before you misalign you can survive the nanobots.
Re:AGI Simulator
by scrumper on Wednesday July 12, 2023 @11:47AM
(#36696357)
What is "AI Wealth" here, where's it measured? Human wealth I get, it's the money stat. Human knowledge is the book. AI IQ is self explanatory. Population, GPUs, same. So which one is AI wealth?
Re:AGI Simulator
by mkaic on Wednesday July 12, 2023 @11:51AM
(#36696441)
There's only one wealth and one knowledge coffer. Both AI and humanity have their own wealth and knowledge production rates linked to the number of GPUs/IQ and population, respectively.
I don't think there's AI knowledge production actually, or at least that's why I assumed since we don't have a way to improve it with research unlike everything else.
Re:AGI Simulator
by shmapf on Wednesday July 12, 2023 @01:07PM
(#36697880)
I don't understand why it resets at universal jump engine. Is that the end of the game, or is there some way to end up in a different state? I got to the point where that was the only option left.
Got hit by the virus.
Maxed alignment after that.
Now in 2066 and waiting for BrainLink
I make "AI Simulator" games, where you use AI to run simulations of games automatically.
https://ai-simulator.com/
[1] http://www.thelastquestion.net/
Second run ('cause I clicked the universal jump engine by accident): I deliberately hold off on AI upgrades unless I can keep alignment high. At 2080, AI is out-earning humanity but very stupid and humanity is only at 586 million.
(Also, why does the population start off so low? Do you know something that I don't?)
yeah a warning about 'universal jump engine' would be good.
I enjoyed the Paperclip thing, this is similar.