I think there’s an excellent case to be made for tech employees who are convinced the singularity is near, a transferable preservation in case a loved one goes terminal in the next few years is a good bet when that cost is a fraction of their annual salary.
romeostevensit
Karma: 11,239
Was able to reproduce in a chrome incognito window but worked in regular chrome and firefox. I didn’t want to paste in a bunch of ai text and the formatting breaks as well
I think it is has been net positive for people to begin having earnest conversations about their real reasons for things in public rather than filter everything through a particular discourse norms filter that in practice means people don’t share their real reasons.
this should be a post not just shortform?
Claude recommended that VCG as a kind of inverse Shapley often works better in the real world by encouraging truth telling. As a bonus, it handles damages (as opposed to benefits) slightly more easily.
Best title I’ve read in a long time.
In America, the credibility of the Common Law suffered a decisive blow around the time of the Civil War, when it failed to address the issue of slavery through legal mechanisms, and Americans resorted to war to settle their dispute. This was a delegitimating crisis, but it took a generation for America’s governance institutions to be captured by a new antinomian ideology annealed by the war. Oliver Wendell Holmes Jr., whose major work begins with The Common Law in 1881, argued that law is prediction of what courts will do, not discovery of pre-existing principles. His Legal Realist heirs in the 1920s and 1930s completed the displacement of Common Law reasoning.[4]The result was Pragmatism:[5]a framework that retained the forms of lawful governance while abandoning the principle that law is discovered rather than made anew each time a court sits.[6]
this is a very good/orienting paragraph.
those AI researchers would leave and go to other companies.
small caveat that I believe it would be positive to concentrate researchers in fewer companies.
Completely inviolable hard lines expose one to cheap blackmail attacks.
Edit: I was sad when I predicted disagreement votes for this and more sad when they manifested.
thank you for not taking the paycheck.
You quote the long list of things he tried that failed then have a just so story about what he has found that somewhat worked. Outlier analysis isn’t a good choice for the intuition pump you’re trying to share in the first place.
in the last 5 months I have learned some things about how I engage in truth seeking with people and have updated in this direction, though still think there are important cases of my original point*. I think learning to distinguish them better has been and continues to be high value for me.
*specifically, that people will de facto prioritize their feelings of discomfort in an argument from consequences way even when the outcomes are morally fraught.
Agree and I think pervasive low morale is causing much of the split between material conditions being fine and people reporting that they feel they are doing poorly. Effort can get you stasis, but it doesn’t feel like it can get you meaningful level ups (for many people, myself included often).
Not meta level helpful but I always cede ground and blacklist the people involved.
The relationship is much more complicated than that. The dog has superpowers and often knows important things that you are getting wrong.
Hot house orchid types with neuroticism and health problems get totally fucked if they don’t have strong familial support IME.
Backing selection effects out of data is a notoriously expensive operation without guarantees of convergence to the true distribution afaik.
My current understanding is that a good way to prevent the sloppification and sycophantic attractors are to not talk directly with the base assistant persona but instead have two higher quality personas talk about your prompt. I’m hoping to see more experimentation with this so I don’t have to build the whole thing myself (ideal version is separate API calls with separate rags and separate parameters).
(The following is an exercise in testing LLM creative writing prompts. Hence why it is not a full post. Text almost entirely generated by Claude)
The Breakfast Question
The question spread the way questions do — innocuously, tucked inside a philosophy blog post nobody would admit to reading. By Tuesday it was a tweet. By Thursday it was on bus shelter ads in four cities, though nobody could say who bought the space.
Most people died mid-thought.
Dr. Miriam Osei was the first epidemiologist to notice the pattern in the bodies. They weren’t sick. They weren’t poisoned. Their faces held a kind of arrested expression — not pain, exactly, but the particular tension of a mind pulling in two directions. She described it in her notes as the look of a door being opened from both sides simultaneously. The neuroscientists came next. The mechanism, they determined, was a prion-adjacent structure that assembled itself in the default mode network — the seat of simulation, projection, retrospection. Of imagined alternatives. It lay dormant until the host attempted to model a world that was not the world. Then it folded, catastrophically, taking everything nearby with it. The virus was, in this sense, extraordinarily precise. It had no interest in memory. It had no interest in regret per se, or hope, or planning. It targeted specifically the act of forking — of holding two timelines simultaneously, comparing them, asking which was better. The question was simply the key that fit the lock.
The survivors organized quickly, as survivors do. Manny had never been a philosophical man. He woke each morning and ate whatever was in the refrigerator without ceremony, without considering whether he should have bought different groceries, without projecting forward to lunch. When the question reached him — his cousin read it aloud at the table, and then his cousin’s face did the thing — Manny had simply blinked. “I ate eggs”, he thought. They were fine. His prions had nothing to grip. He became, within six months, a mayor of sorts. The community he led was not stupid. This was the thing the surviving intellectuals struggled to articulate to each other, in the first weeks, before most of them died reaching for counterfactual comparisons mid-sentence: the survivors were not cognitively impaired. They could reason. They could infer. They could solve problems. They simply could not — constitutionally, neurologically, perhaps spiritually — wonder what would have happened if. A child named Soo-Ah, age seven, turned out to be a gifted engineer. She could look at a broken thing and understand immediately what it needed. She had no capacity to imagine a version of the thing that had never broken, no grief for the pristine alternative. She simply fixed what was in front of her. Manny watched her work one afternoon and felt something he couldn’t name. It was close to awe.
The philosopher who had written the original blog post survived too. This was either ironic or fitting, depending on your capacity for that distinction — which, increasingly, was not a capacity anyone had. Her name was Dr. Lena Voss, and she had survived because she had been, for the previous four years, working on a paper arguing that counterfactual reasoning was not native to human cognition but was instead a learned and fragile heuristic — a cognitive prosthetic that cultures had developed and individuals had adopted, unnaturally, like shoes. In arguing this, she had trained herself to notice the exact moment she began to fork a timeline. And in noticing it, she had learned, slowly, to stop. She had not intended this as prophylaxis. She had not imagined — there, that, she would catch herself, stop — she had not expected her work to be anything other than a small contribution to embodied cognition literature. She lived because she had learned to stay inside the actual. She spent her remaining years teaching others to do the same, though the ones who most needed it were already dead.
What was lost was enormous and everyone knew it without being able to articulate it precisely. Not imagination — the survivors dreamed richly, planned, invented. What was lost was the particular human ache of the other branch — the version of yourself who turned left, who said yes, who ate the eggs or didn’t. The mourning of paths not taken. The comfort of a road untraveled that you could return to, mentally, in the dark. Soo-Ah fixed things beautifully and never wondered what they might have been. Manny made good decisions and never second-guessed them. They were, in some technical sense, happier. Whether this was a mercy or the final joke of the virus — whether the question had, in answering itself, answered more than it asked — was a thought that died with those who could have thought it.