But fighting for public epistemology is a long battle; it makes more sense if you have _time_ for it to pay off. Back in the late 'aughts and early 'tens, it looked like we had time. We had these abstract philosophical arguments for worrying about AI, but no one really talked about _timelines_. I believed the Singularity was going to happen in the 21st century, but it felt like something to expect in the _second_ half of the 21st century.
-Now it looks like we have—less time? Not just tautologically because time has passed (the 21st century is one-fifth over—closer to a quarter over), but because of new information from the visible results of the deep learning revolution.[^second-half] Yudkowsky seemed particularly [spooked by AlphaGo](https://www.lesswrong.com/posts/7MCqRnZzvszsxgtJi/christiano-cotra-and-yudkowsky-on-ai-progress?commentId=gQzA8a989ZyGvhWv2) [and AlphaZero](https://intelligence.org/2017/10/20/alphago/) in 2016–2017.
+Now it looks like we have—less time? Not just tautologically because time has passed (the 21st century is one-fifth over—closer to a quarter over), but because of new information from the visible results of the deep learning revolution.[^second-half] Yudkowsky seemed particularly [spooked by AlphaGo](https://www.lesswrong.com/posts/7MCqRnZzvszsxgtJi/christiano-cotra-and-yudkowsky-on-ai-progress?commentId=gQzA8a989ZyGvhWv2) [and AlphaZero](https://intelligence.org/2017/10/20/alphago/) in 2016–2017, not because superhuman board game players were dangerous, but because of what it implied about the universe of algorithms.
+
+There had been a post in the Sequences that made fun of "the people who just want to build a really big neural net." These days, it's increasingly looking like just building a really big neural net ... [actually works](https://www.gwern.net/Scaling-hypothesis)?—which is bad news; if it's "easy" for non-scientific-genius engineering talent to shovel large amounts of compute into the birth of powerful minds that we don't understand and don't know how to control, then it would seem that the world is soon to pass outside of our understanding and control.
[^second-half]: In an unfinished slice-of-life short story I started writing _circa_ 2010, my protagonist (a supermarket employee resenting his job while thinking high-minded thoughts about rationality and the universe) speculates about "a threshold of economic efficiency beyond which nothing human could survive" being a tighter bound on future history than physical limits (like the heat death of the universe), and comments that "it imposes a sense of urgency to suddenly be faced with the fabric of your existence coming apart in ninety years rather than 10<sup>90</sup>."
My AlphaGo moment was 5 January 2021, when OpenAI released [DALL-E](https://openai.com/blog/dall-e/) (by far the most significant news story of that week in January 2021). Previous AI milestones, like GANs for a _fixed_ image class, were easier to dismiss as clever statistical tricks. If you have thousands of photographs of people's faces, I didn't feel surprised that some clever algorithm could "learn the distribution" and spit out another sample; I don't know the _details_, but it doesn't seem like scary "understanding." DALL-E's ability to _combine_ concepts—responding to "an armchair in the shape of an avacado" as a novel text prompt, rather than already having thousands of examples of avacado-chairs and just spitting out another one of those—viscerally seemed more like "real" creativity to me, something qualitatively new and scary.[^qualitatively-new]
-[^qualitatively-new]: By mid-2022, DALL-E 2 and Midjourney and Stable Diffusion were generating much better pictures, but that wasn't surprising. Seeing AI being able to do a thing _at all_ is the model update; AI being able to do the thing much better 18 months later feels "priced in."
+[^qualitatively-new]: By mid-2022, DALL-E 2 and Midjourney and Stable Diffusion were generating much better pictures, but that wasn't surprising. Seeing AI being able to do a thing at all is the model update; AI being able to do the thing much better 18 months later feels "priced in."
[As recently as 2020, I had been daydreaming about](/2020/Aug/memento-mori/#if-we-even-have-enough-time) working at an embryo selection company (if they needed programmers—but everyone needs programmers, these days), and having that be my altruistic[^eugenics-altruism] contribution to the great common task. Existing companies working on embryo selection [boringly](https://archive.is/tXNbU) [market](https://archive.is/HwokV) their services as being about promoting health, but [polygenic scores should work as well for maximizing IQ as they do for minimizing cancer risk](https://www.gwern.net/Embryo-selection).[^polygenic-score] Making smarter people would be a transhumanist good in its own right, and [having smarter biological humans around at the time of our civilization's AI transition](https://www.lesswrong.com/posts/2KNN9WPcyto7QH9pi/this-failing-earth) would give us a better shot at having it go well.[^ai-transition-go-well]
But pushing on embryo selection only makes sense as an intervention for optimizing the future if AI timelines are sufficiently long, and the breathtaking pace (or too-fast-to-even-take-a-breath pace) of the deep learning revolution is so much faster than the pace of human generations, that it's starting to look unlikely that we'll get that much time. If our genetically uplifted children would need at least twenty years to grow up to be productive alignment researchers, but unaligned AI is [on track to end the world in twenty years](https://www.lesswrong.com/posts/AfH2oPHCApdKicM4m/two-year-update-on-my-personal-ai-timelines), we would need to start having those children _now_ in order for them to make any difference at all.
-[It's ironic that "longtermism" got traction as the word for the "EA" cause area of benefitting the far future](https://applieddivinitystudies.com/longtermism-irony/), because the decision-relevant beliefs of most of the people who think about the far future, end up working out to extreme _short_-termism. Common-sense longtermism—a longtermism that assumed there's still going to be a world of recognizable humans in 2123—_would_ care about eugenics, and would be willing to absorb political costs today in order to fight for a saner future. The story of humanity would not have gone _better_ if Galileo had declined to publish his works for fear of the Inquisition.
+[It's ironic that "longtermism" got traction as the word for the "EA" cause area of benefitting the far future](https://applieddivinitystudies.com/longtermism-irony/), because the decision-relevant beliefs of most of the people who think about the far future, end up working out to extreme _short_-termism. Common-sense longtermism—a longtermism that assumed there's still going to be a recognizable world of humans in 2123—_would_ care about eugenics, and would be willing to absorb political costs today in order to fight for a saner future. The story of humanity would not have gone _better_ if Galileo had declined to publish for pre-emptive fear of the Inquisition.
-But if you think the only hope for there _being_ a future flows through maintaining influence over what large tech companies are doing as they build transformative AI, declining to contradict the state religion makes more sense—if you don't have _time_ to win a culture war, because you need to grab hold of the Singularity (or perform a [pivotal act](https://arbital.com/p/pivotal/) to prevent it) _now_.
+But if you think the only hope for there _being_ a future flows through maintaining influence over what large tech companies are doing as they build transformative AI, declining to contradict the state religion makes more sense—if you don't have _time_ to win a culture war, because you need to grab hold of the Singularity (or perform a [pivotal act](https://arbital.com/p/pivotal/) to prevent it) _now_. If the progressive machine marks you as a transphobic bigot, the machine's functionaries at OpenAI or Meta AI Research are less likely to listen to you when you explain why their safety plan won't work (or why they should have a safety plan at all).
(I remarked to "Wilhelm" in June 2022 that DeepMind changing its Twitter avatar to a rainbow variant of their logo for Pride month was a bad sign.)
+So isn't there a story here where I'm the villain, willfully damaging humanity's chances of survival by picking unimportant culture-war fights in the xrisk-reduction social sphere, when _I know_ that the sphere needs to keep its nose clean in the eyes of the progressive egregore? _That's_ why Yudkowsky said the arguably-technically-misleading things he said about my Something to Protect: he _had_ to, to keep our nose clean. The people paying attention to contemporary politics don't know what I know, and can't usefully be told. Isn't it better for humanity if my meager talents are allocated to making AI go well? Don't I have a responsibility to fall in line and take one for the team—if the world is at stake?
-[...]
+As usual, the Yudkowsky of 2009 has me covered. In his short story ["The Sword of Good"](https://www.yudkowsky.net/other/fiction/the-sword-of-good), our protagonist Hirou wonders why the powerful wizard Dolf lets other party members risk themselves fighting, when Dolf could have protected them:
-> [_Perhaps_, replied the cold logic. _If the world were at stake._
+> _Because Dolf was more important, and if he exposed himself to all the risk every time, he might eventually be injured_, Hirou's logical mind completed the thought. _Lower risk, but higher stakes. Cold but necessary–_
+>
+> _But would you_, said another part of his mind, _would you, Hirou, let your friends walk before you and fight, and occasionally die, if you_ knew _that you yourself were stronger and able to protect them? Would you be able to stop yourself from stepping in front?_
+>
+> _Perhaps_, replied the cold logic. _If the world were at stake._
>
-> _Perhaps_, echoed the other part of himself, _but that is not what was actually happening._](https://www.yudkowsky.net/other/fiction/the-sword-of-good)
+> _Perhaps_, echoed the other part of himself, _but that is not what was actually happening._
-[TODO: social justice and defying threats
-
- * There's _no story_ in which misleading people about transgender is on Yudkowsky's critical path for shaping the intelligence explosion. _I'd_ prefer him to have free speech, but if he can't afford to be honest about things he already got right in 2009, he could just—not bring up the topic!
+That is, there's _no story_ under which misleading people about trans issues is on Yudkowsky's critical path for shaping the intelligence explosion. _I'd_ prefer him to have free speech, but if _he_ thinks he can't afford to be honest about things he [_already_ got right in 2009](https://www.lesswrong.com/posts/QZs4vkC7cbyjL9XA9/changing-emotions), he could just—not issue pronouncements on topics where he intends to _ignore counterarguments on political grounds!_
+
+In [a Twitter discussion about why not to trust organizations that refuse to explain their reasoning, Yudkowsky wrote](https://twitter.com/esyudkowsky/status/1374161729073020937):
+
+> Having some things you say "no comment" to, is not at _all_ the same phenomenon as being an organization that issues Pronouncements. There are a _lot_ of good reasons to have "no comments" about things. Anybody who tells you otherwise has no life experience, or is lying.
-https://twitter.com/esyudkowsky/status/1374161729073020937
-> Also: Having some things you say "no comment" to, is not at *all* the same phenomenon as being an organization that issues Pronouncements. There are a *lot* of good reasons to have "no comments" about things. Anybody who tells you otherwise has no life experience, or is lying.
+Sure. But if that's your story, I think you need to _actually not comment_. ["[A]t least 20% of the ones with penises are actually women"](https://www.facebook.com/yudkowsky/posts/10154078468809228) is _not "no comment"._ ["[Y]ou're not standing in defense of truth if you insist on a word, brought explicitly into question, being used with some particular meaning"](https://twitter.com/ESYudkowsky/status/1067198993485058048) is _not "no comment"_. We did get a clarification on that one—but then, within a matter of months, he turned around and came back with his "simplest and best proposal" about how the "important things [...] would be all the things [he's] read [...] from human beings who are people—describing reasons someone does not like to be tossed into a Male Bucket or Female Bucket, as it would be assigned by their birth certificate", _which is also not "no comment."_
- * I can totally cooperate with censorship that doesn't actively intefere with my battle! I agree that there are plenty of times in life where you need to say "No comment." But if that's the play you want to make, you have to actually _not comment_. "20% of the ones with penises" is no "No comment"! "You're not standing in defense of truth" is not "No comment"! "The simplest and best proposal" is not "No comment"!
+[TODO: defying threats, cont'd—
* I don't pick fights with Paul Christiano, because Paul Christiano doesn't take a shit on my Something to Protect, because Paul Christiano isn't trying to be a religious leader. If he has opinions about transgenderism, we don't know about them.
marked TODO blocks—
✓ AlphaGo seemed deeper [pt. 5]
- "Agreeing with Stalin" recap intro [pt. 5]
-_ social justice and defying threats [pt. 5]
+- social justice and defying threats [pt. 5]
_ scuffle on "Yes Requires the Possibility" [pt. 4]
_ confronting Olivia [pt. 2]
_ "Lesswrong.com is dead to me" [pt. 4]
With internet available—
+_ real-name blog post: jr. member of save/destroy/take-over the world conspiracy
+_ Sequences post making fun of "just make a really big neural net"
_ DeepMind June 2022 Twitter archive?
_ January 2021 Capitol riot
_ No Fire Alarm
_ better introduction of S.K.
_ risk of people bouncing off progressivism
_ an AGP teen boy could at least consent to transition, and make plans based on knowing what the thing is (you'd actually want to go through a little bit of male puberty)
+_ "tossed in a bucket" is ignoring advice from Sept. 2022 clarification to be clear about the type distinction
_ explicitly explain "Hill of Validity" title inversion
_ figure out full timeline of which of my Less Wrong posts to mention
_ update "80,000 words" refs with the near-final wordcount
projects / Ultimately_Untrue_Thought.git / commitdiff