Max was a gainsmaxxer. His singular goal in life was to maximize his muscle mass by optimizing mechanical tension and muscle-protein synthesis. He went to the gym every day in hopes of maximizing his gains.
Every day, he drove home from the gym, because walking home would increase his non-exercise activity thermogenesis, which might create a caloric deficit and kill his gains.
One day, he drove home and found that the elevator was broken. Climbing the single fight of stairs to his apartment would activate his glutes without providing enough tension to stimulate meaningful hypertrophy. Junk volume might kill his gains, so he walked back to the parking lot and slept in his car. (He made sure not to sleep on his side, because shoulder compression might kill his gains.) After sleeping for exactly 9 hours to maintain maximum growth hormone pulsatility, he drove back to the gym.
After Max had completed his third warm-up set for lateral raises, Max was approached by a pretty woman. She said, "Hey big guy, can you help me re-rack some of these weights?"
Max froze. He normally avoided talking to attractive women, because dopamine volatility might kill his gains. But as the woman smiled at him, he realized that he had spent 3000 hours in the gym preparing for this moment, when a woman would look at him, notice the size of his delts, and offer him the chance to help her by lifting dumbbells.
Max said, "No. Lifting those weights would cause grip fatigue that isn't programmed into my mesocycle. Spending my limited recovery resources on non-hypertropic labor might kill my gains."
As he exited the gym, a bearded man in a gold hat approached him. "Hello," said the man. "My name is Eliezer Yudkowsky."
"What do you want?" asked Max.
"I'm concerned about AI risk. If we build systems that are more intelligent than humans and fail to properly align them with human values, they may pursue goals that destroy us, even if they have terminal goals that seem naively aligned with what we want. For example, your body is made of atoms that the AI might think are more useful for other purposes."
Max felt his cortisol levels rising, but he couldn't ignore what he was hearing. That night, he began to research AI alignment, and became more and more worried about the possibility of AI risk. He read papers by Apollo Research and began working on propensity evals in his spare time, then applied to MATS.
When submitting his application, the final question asked "what made you choose to work in this cause area?" Max typed: "Misaligned AI might kill my gains."
Impressive post. I’m sure the slang will hit hard with the youth.
I will still not be valuing the soil nematodes, shrimp, or the hypothetical 10^54 future minds as more important than the pepperoni pizza I am currently eating.
For those interested in 'maxxing' through a more personalized lens, I set out a pragmatic meta-ethics, 'Heirs of Life-Years' (happy to share WIP long form manuscript), starting in the below post:
Basically you can ground ethics with a count and a clock...but there's no view from nowhere to tell anyone a 'correct' counting procedure. Thus, in practice, to follow through on:
|The good is good, so you should bring about more of it.
...pragmatically, *you have to specify what counting procedure you're using to count e.g. "good lives" in a room*. More importantly - as a finite being, it is impossible to care about all lives equally (H3). This leads to my modernized definition of the word "heirs" - the present and future lives for which an agent takes primary and enduring ethical responsibility.
- Similar to the argument above, we get a calculable version of the Rawlsian Veil - the Veil of Future Heirs, generalizing democratic and evolutionary principles:
- As I show in the full manuscript, but can be easily seen with a few minutes of thought - we recover a pragmatic form of universalism, by noting that two *currently* disjoint heir definitions or communities, under assumptions of minimal mixing and non-ruin (e..g both not going extinct), on long horizons align due to both heir-sets covering 'all' of e.g. humanity.
deontcels seething rn
Finally you are speaking a language that based gen z goodmaxxers like me understands!
I don't think many of your readers are gen Alpha though I may be wrong. I appreciate the sentiment regardless.
Max was a gainsmaxxer. His singular goal in life was to maximize his muscle mass by optimizing mechanical tension and muscle-protein synthesis. He went to the gym every day in hopes of maximizing his gains.
Every day, he drove home from the gym, because walking home would increase his non-exercise activity thermogenesis, which might create a caloric deficit and kill his gains.
One day, he drove home and found that the elevator was broken. Climbing the single fight of stairs to his apartment would activate his glutes without providing enough tension to stimulate meaningful hypertrophy. Junk volume might kill his gains, so he walked back to the parking lot and slept in his car. (He made sure not to sleep on his side, because shoulder compression might kill his gains.) After sleeping for exactly 9 hours to maintain maximum growth hormone pulsatility, he drove back to the gym.
After Max had completed his third warm-up set for lateral raises, Max was approached by a pretty woman. She said, "Hey big guy, can you help me re-rack some of these weights?"
Max froze. He normally avoided talking to attractive women, because dopamine volatility might kill his gains. But as the woman smiled at him, he realized that he had spent 3000 hours in the gym preparing for this moment, when a woman would look at him, notice the size of his delts, and offer him the chance to help her by lifting dumbbells.
Max said, "No. Lifting those weights would cause grip fatigue that isn't programmed into my mesocycle. Spending my limited recovery resources on non-hypertropic labor might kill my gains."
As he exited the gym, a bearded man in a gold hat approached him. "Hello," said the man. "My name is Eliezer Yudkowsky."
"What do you want?" asked Max.
"I'm concerned about AI risk. If we build systems that are more intelligent than humans and fail to properly align them with human values, they may pursue goals that destroy us, even if they have terminal goals that seem naively aligned with what we want. For example, your body is made of atoms that the AI might think are more useful for other purposes."
Max felt his cortisol levels rising, but he couldn't ignore what he was hearing. That night, he began to research AI alignment, and became more and more worried about the possibility of AI risk. He read papers by Apollo Research and began working on propensity evals in his spare time, then applied to MATS.
When submitting his application, the final question asked "what made you choose to work in this cause area?" Max typed: "Misaligned AI might kill my gains."
You just absolutely mogged ACX. Keep on blogmaxxing!
-- Eli Goldfine (b. 2012)
Such a great inspiring article. Goodmaxxing it will be for no matter what !!!
If you think if making people feel temporary pleasure is good, wait until you find out about immortal souls
Pretty sure Matthew already believes in immortal souls? (Matthew 10:28)
Haha well played
Impressive post. I’m sure the slang will hit hard with the youth.
I will still not be valuing the soil nematodes, shrimp, or the hypothetical 10^54 future minds as more important than the pepperoni pizza I am currently eating.
The goodmaxxers will moralmog you!
Nice rage bait, I rate it 6-7 out of 10.
Absolutely based woke and new bee.
It's wrong to lie.
For those interested in 'maxxing' through a more personalized lens, I set out a pragmatic meta-ethics, 'Heirs of Life-Years' (happy to share WIP long form manuscript), starting in the below post:
https://pragmaticfutures.substack.com/p/heirs-of-life-years-a-meta-ethics
Basically you can ground ethics with a count and a clock...but there's no view from nowhere to tell anyone a 'correct' counting procedure. Thus, in practice, to follow through on:
|The good is good, so you should bring about more of it.
...pragmatically, *you have to specify what counting procedure you're using to count e.g. "good lives" in a room*. More importantly - as a finite being, it is impossible to care about all lives equally (H3). This leads to my modernized definition of the word "heirs" - the present and future lives for which an agent takes primary and enduring ethical responsibility.
- Similar to the argument above, we get a calculable version of the Rawlsian Veil - the Veil of Future Heirs, generalizing democratic and evolutionary principles:
https://pragmaticfutures.substack.com/p/the-veil-of-future-heirs-holys-excerpt
- As I show in the full manuscript, but can be easily seen with a few minutes of thought - we recover a pragmatic form of universalism, by noting that two *currently* disjoint heir definitions or communities, under assumptions of minimal mixing and non-ruin (e..g both not going extinct), on long horizons align due to both heir-sets covering 'all' of e.g. humanity.