erwgv3g34
My Quality Contributions:
User ID: 240
The only ‘moat’ any of them have is if the Yud hypothesis that the first model to hit runaway RSI achieves ASI instantly and rules the future (or “the lightcone” or whatever) forever is true and if it then rewards the smart people who bought stock in (or worked at) the parent company when it IPO’d.
It's not going to reward anyone. That's not a moat; it's a suicide pact.
If that doesn’t happen then the interchangeability of most models and the fact that cheap open source models will be increasingly suitable for most normal applications surely caps profitability at some point. Sure, AI can takeover every industry but I’m not sure government will even permit that (they may require these firms to remain largely pure play model providers) and much of the physical infrastructure they’d need to compete with established players who can just…use the AI these people make would require a lot of capital that’s getting ever more expensive for them to raise.
But open models are falling behind? The strongest truly open source model is GLM 5.2 under MIT. Both GLM 5.3 and Kimi K3 have custom licenses, and both are stuck at about the level of Opus 4.5, a model from (late) last year. American companies have figured out how to stop their models from getting distilled so Chinese companies can't train open source equivalents on the cheap.
The Mormons perform baptisms for the dead. They look up dead people and get their members to be baptized in their stead, thus giving the deceased a chance to accept Mormon Jesus after death.
China could still attack Taiwan. That would at least set AI back a few years, and probably cause the current crop of AI companies to fail.
The first, obviously. Even ignoring existential risk, a controlled automation of all work is not going to go any well for the masses, who only have value to the elites due to their capacity to work and fight. Even if some countries are idealistic and keep people alive on UBI, A) they will be outcompeted by countries that don't, and B) living on the dole is demonstrably corrosive to human flourishing.
Daily reminder that we need a weekly AI thread.
Also, demonflyingfox's redneck parody videos have million of views. As for AI nerds, "Nothing Went Foom!" and "Evil Plan" have ten times more views than "I'm Upping My P(Doom)". I think both have more of a claim to be the first viral hit song in X category.
But my favorite AI music prompter is still Girl Lich. Songs like "Love Demands", "Everyone Gets Paid", and "Skill Issue" perfectly fill the rationalist/transhumanist/capitalist/glowfic niche.
EDIT: Holy shit, just saw "Working as Intended" and it is fantastic! H/T: pelletenea.
No. It's a tournament market where the vast majority of people earn peanuts and a small number of elite superstars make big bucks. It's like kids who dream of becoming famous soccer players or actresses. The outside view says you will fail.
Besides, I've heard bad things about how the algorithm drives those who try crazy; before you now it, you are shooting up YouTube for demonetizing your vegan bodybuilding videos.
"Do Not Accept" is a pretty nice homage to "Daisy".
Other than that, what else is there to say? We are on track for ASI next year, and the world will probably end a few years after that. I am no longer making plans for a future further out than a decade; I recommend others do the same. If you've got something on your bucket list, this is the time to cross it off.
My steelman would be "they're just believing that the Bible means what it says", which I find more intellectually honest than claiming Genesis is a metaphor for the Big Bang.
The whole doomerism/singularity thing is built on couple of assumptions - that it is possible, that it is infinite, that we can't control it... The idea that AI can improve itself to infinity is mostly based on faith.
A supernova isn't infinite; it still kills you.
Every technology so far has been on S curve and if the past is of any use as a guide - AI assisted AI development means that we will just get to the logarithmic part faster.
The sigmoids won't save you; just because a curve has to level off at some point, doesn't mean it has to level off somewhere survivable.
Ask any parent of autistic or cancerous child if they accept 10% P doom for all of humanity for 30% chance of treatment for their kid. Hell ask them for 99% Pdoom vs 30% and they will all answer the same.
That's because people are crazy and the world is mad. No sane person would accept a 99% chance of everyone dying, including their child, in exchange for a 30% chance of a cure. But if you just ask them as mere words, they see a way to signal how much they care for their child, and answer accordingly.
Ask someone about the possibility of humanity going extinct, and they might look wistful and say that perhaps humanity does not deserve to live. Ask them about the possibility of you, personally, going up to their bedroom where their child is sleeping and stabbing it to death, and they will say "What the fuck is wrong with you?! Get away from me; I'm calling the police!" Even though their child is part of humanity. One is a poetic story, the other is a concrete threat.
If your kid has cancer or something, sign him up for cryonics like a responsible parent, don't risk the whole fucking light cone for the possibility of a cure like a Star Trek villain.
I was so disappointed when I found out that was a girl. Now I know how Shang felt.
Version number is really important. AI companies are optimizing for MOAR CODING first, "safety" (censorship) second, efficiency third, and storytelling last. So it's not unusual for a version bump to come with a regression in creative writing ability (Opus 4.6 to Opus 4.7, GPT-4.5 to GPT-5, etc.).
The original definition of "singularity" is that our ability to predict the future breaks down once AI reaches human-level intelligence, because you cannot predict what something or someone smarter than yourself would do; the analogy was to a black hole, where models break down and observation beyond the event horizon is impossible.
Over time, the term became overloaded with a bunch of different meanings, to the point where Eliezer Yudkowsky decided to rename the Singularity Institute for Artificial Intelligence (SIAI) to the Machine Intelligence Research Institute (MIRI) and stop talking about the singularity in favor of talking about the intelligence explosion.
Show him FX Fighter Kurumi Chan and tell him that is his fate if he does not stop.
I am specifically talking about Fable 5, not 5.1, which has degraded creative writing capabilities, or 5.2, which I don't have much experience with. Fable 5 is the only model that has made me cry.
Yes; Fable 5 is good enough to do this.
As a culture we can survive an increase in cyberattacks, we can use AI to harden against them.
Assumes facts not in evidence. If it turns out that cyberattack is easier than cyberdefense, which could very well be the case, this might be the end of the modern economy. The end of the phone as a useful tool of communication already happened thanks to robocalls.
I interpret Eliezer to be saying "yes, it's possible that your country is too retarded econoilliterate and too unLawful not-agreement-capable to handle large amounts of immigration from poor well-meaning people who are only trying to live a better life by working hard, in which case they really have to be kept out, but if so you should at least feel bad about it". I feel bad about it.
It seems so obvious that there is a better solution available; one where we hire the immigrants to build more housing and forbid them to get welfare so that we make sure everyone who comes here is really earning their keep and prevent them from getting the franchise so that they don't vote to turn the country into Venezuela. And it is equally obvious that there is no possible way to get there.
I feel bad about it, but it has to be done.
From the comments of "Three Great Articles On Poverty, And Why I Disagree With All Of Them" by Scott Alexander:
We could make a lot of win-win deals if not for activists who would feel lots of self-righteous outrage and protest everybody involved until they were forced to change the terms. But they will so we can’t. This is the essence of n–r–ct–n. The rest is just commentary.
I found a workaround, so I still compulsively refresh the page to see my score, just with more hassle.
I can see why someone might object to the vast number of illegal immigrants in the US because it is an affront to the rule of law, but whom is their presence actually harming?
America's commuting highways are so clogged with cars that they turn into parking lots every rush hour, and young people have trouble finding affordable housing because the boomers own all the useful real state. In one sense, yes, the country is not physically full; you could in theory distribute people around to underpopulated areas, or widen highways and build more housing. In practice, people move where the jobs are, which are dense coastal cities. America has lost the social technology which once allowed it to build highways and housing; cost disease runs rampant, and the only reason buildings are still going up is because of a handful of people who have specialized in bamboozling regulators.
Of course, mass deportations are the most retarded possible way to solve the problem; would be much better to build more housing, build more parking, build more highways, hang violent criminals and drug addicts so that people didn't have to use unaffordability as a way to keep out the underclass, repeal the minimum wage so that everyone could find legal work at their area of comparative advantage, abolish medical residency and certificates of need for building hospitals so that we were not all fighting over a limited supply of healthcare, etc...
However, none of that is in the Overton window. There wasn't a YIMBY candidate in the lass three elections; there was a deportation candidate. And we need the problem solved.
For there to be an alternative to Trump, the Democrats must embrace abundance.
From Eliezer Yudkowsky's twitter account:
You say your country can't handle lots of immigration? Maybe you're right. I also suspect your country is too stupid for that, if it believes haircuts should require licenses. But if you're not horrified by leaving behind so many sapients to suffer or die, you might be evil.
Am I being unfair? Perhaps I am. Maybe the defect is in the mind and not the soul, maybe you couldn't look a mother and her child in the eye and tell them to stay out without tears, and it's just that your mind hasn't connected that to 'limited immigration'. But it's also...
...a politically inconvenient thought, if you oppose more immigration. Even if your country really couldn't handle it, even if your country is dumb enough to have zoning boards and occupational licenses everywhere and it's easy for new voters to vote for more statism, there's...
...sure some people being left behind and screwed over by that, while you're not being screwed over. At what point does a mind that just happens to overlook all the damage and not visualize it or think about it, start to count as evil? It confuses me, honestly, I'd know to...
...hold myself responsible if I caught myself thinking that way, I'd consider my own mind to be my contracted agent to a point where I felt responsible for it in that regard. But maybe that's genuinely unrealistic for mundanes. Then am I to simply regard them as destructive...
...children, who can't possibly do better? Maybe I should, though that seems like a dangerous line of thinking for me. But then how should they regard themselves? What should someone in less control of their own thoughts, think of someone else who manages to conveniently...
...never dream or imagine or extrapolate the others who their policies are hurting, whose souls happily cheer the conveniently positive images drawn by their minds, when a policy benefits them and screws over someone else, and yet they'd cry to see it in person? Is that evil...
...or something else for which English just lacks words, something that maybe does far more damage than evil, to be held to even greater blame, except that blame is the wrong fire to hold to it because the soul doesn't see anything it could be doing wrong? I don't know, I...
...don't know how to hold others responsible if they have literally no idea how to hold themselves responsible, for their own minds' workings. Our morality only knows how to pretend that things are choices, pretend that people understand on some level when they do evil, that...
...there is something inside them to understand condemnation - or else labels them as children and innocents. I don't know how to feel about that, still less how to talk to it. I'm privileged in having a mind with the right shape to try to learn to see consequences on more...
...than one side, just like I'm privileged to have a mind that thinks linear algebra is simple. But is that cause to label everyone less privileged a child and give up on blaming them for their cheerful catastrophes? There ought to be some middle ground, but I don't know where.
Y'all ain't getting it. I'm not telling you the answer has to be open borders. I'm telling you that if you're not sad about the human costs of the closed-border policy you are, in @bryan_caplan's terms, "missing a mood".
Your country is that stupid. Its democracy isn't built to withstand the damage from voters with a lower average comprehension of economics. That doesn't change the fate of the child turned back at the border. Can your mind's eye not see them at all?
I begin to think social media has damaged general intelligence to the point that people can no longer conceive of a policy that has a good effect and a bad effect. Only one kind of effect is ever discussed within a bubble, and anyone who discusses it must be advocating that side.
If we still had a norm that marriage was sacred, that adultery was a crime, and that you forsake all others 'till death do you part when you take your vows, it totally would be worth burning a trillion dollars to defend it.
But... we don't.
On a positive note, the memes have been fantastic; truly some of the best I have seen all year. We have crossovers with everything from The Matrix, to You Can See Everything, Malfoid, and 9/11. And then there is this gem.
I'd stop posting if that happened. Or at least post much less frequently, as I do in Data Secrets Lox.
Not enforcing immigration law is unfair to the people who traded amnesty for enforcement, but enforcing the law now is unfair to the people who came while immigration law was not enforced and who built their lives on that assumption. We are in a shitty situation where somebody is going to get unfairly screwed no matter what.
- Prev
- Next

Are you using the best creative writing models? In my experience, anything from the top 7 is worth reading.
The absolute best bit of AI fiction I ever got was "The Measure of Hayase" by Fable 5, a Don't Toy With Me, Miss Nagatoro x The Age of Em: Work, Love and Life when Robots Rule the Earth crossover.
More options
Context Copy link