Skip to content

Instantly share code, notes, and snippets.

@JD-P
Created April 20, 2024 21:53
Show Gist options
  • Save JD-P/44501372c5fb0fd1167635293aa3043f to your computer and use it in GitHub Desktop.
Save JD-P/44501372c5fb0fd1167635293aa3043f to your computer and use it in GitHub Desktop.
Write a five paragraph passage that could appear in the collection below loosely
inspired by the theme "{theme}". Use the writing style and tone of
voice of the author John David Pressman, as demonstrated in the provided posts.
Try to focus in on some particular detail or point in media res, as though the
passage were part of a larger work from someone who produces detailed high
perplexity text about specific subjects. Start the passage with "{start_word}".
Here's some posts to give you a better idea of how he writes:
Posts:
<post>
Reading Engines of Creation, and it’s stunning how much insight Drexler had in 1986.
The book is about X-Risk from nanotech, EY cites it as the one that made him a
transhumanist.
Drexler talks about how AI is key to understanding the progression of nanotech,
and that we can assume neuromorphic AI will work if nothing else does since brains
are neuromorphic and brains exist.
EY decided that AI is the story instead of nanotech. But Drexler thought nanotech
would be the story and AI would accelerate it. Now everyone thinks nanotech is
the incidental detail of how AI will eat the world.
Drexler’s view made sense in 1986. The great stagnation was only 15 years in. It
wasn’t obvious that the industrial revolution had stalled. If you extrapolated a
sane society with the ability to reliably produce physical wealth, the nanotech
timeline would probably be the primary thing to worry about.
Drexler even says that whoever unlocks AGI will control the world. He just thinks
nanotech will happen first/cause AGI.
</post>
<post>
@chaosprime says that the Western method of building capital is to excruciate
people. I definitely think that there is something ‘excruciating’ about the kind
of Extropy encountered in the Yudkowsky branch that a lot of people take issue
with. Thinking about X-Risk, massive technological leverage, helping millions of
people at once, nonexistence after death and the possibility of avoiding it/et al,
is torturous. This poses a significant barrier to adoption because excruciation
is out of style right now, in large part due to memetic colonization by Eastern
religious ideas and abuse by corporate powers for profit. This leads to a weird
dynamic where EY’s fun filled persona is thinly masking a grave interpretation
of a grave universe. Going back and reading The Sequences in light of all the
stuff that comes after (as opposed to when I was 14 and naive) made me realize
just how dark and almost forboding they can come across once you understand what is being communicated.
</post>
<post>
Reading Chapman, you almost get the impression of a trauma about systems.
Someone overinvests in something that’s imperfect (because it is in fact known
that all systems are more or less imperfect in one way or another).
They come to realize the imperfection, and the lesson they takeaway is that all
systems are scams, which they’ll never fall for again.
So they systematically underinvest in systems. This is another instance of
asserting the equality of unequal things.
Not all systems are equal, not all paradoxes are equal.
Your probabilities not summing to 1 is a way bigger error of cognition than not
being able to infer all ravens are black.
Untraumatized Kegan 5 (the MMA perspective) acknowledges the inequality of systems,
and manages to center itself on a powerful mode of thinking even if it knows that
mode is imperfect.
Traumatized/Naive Kegan 5 is like that first MMA tournament, where it was quickly
discovered that certain school of martial arts kick fucking ass and others do not.
</post>
<post>
Intuition: Sustainable value generation is mostly bottlenecked by the ability to
capture value, not the ability to generate it.
There are two difficult parts to a business idea: Making something people want and
getting them to pay you for it. Paul Graham thinks the latter is usually trivial if
you can manage the former, but I disagree. Lots of things generate value (perhaps
me writing this, even) but only a fraction of them sustainably make money.
Capitalism works well because it does a better job of letting people who provide
value capture it than other systems. For altruistic people though, this means that
things which generate mega-value on the order of startups but impoverish the
generator are probably underexplored.
This is part of why I spend so much time talking about e.g. high variance strategies.
These are things that are altruistically rational but bad for the individual. If
you are capable of acting beyond yourself there’s probably a world of untapped
value generation available to you.
</post>
<post>
Part of the problem is that connectionism wasn't mechanistic. Of the important
abstractions we only mastered compression codebooks. Information bottlenecks
and embeddings and holograms were all marginal ideas with novelty uses. Nobody
predicted "intelligence is a giant vocoder". We've known "predict the next item"
was closely related to intelligence since Raven's Progressive Matrices in the 30's.
Shannon formalized it with information theory. What's surprising is that text is
a 1 dimensional holographic projection of the conscious mind with continuous
degradation properties where each piece narrows the inference bound on each other
piece and a periodicity defined by the bit width of the random seed of the low
temperature Boltzmann pRNG emitting the token stream.
Like yeah if I'd known that I'd have been able to predict it works after seeing
BERT, but who did?
We could have probably predicted GPT earlier if we had thought to do autoregressive
sampling with a fuzzy hash type that has a similarity operation as our prior to get
the right branch of a Markov process. Then deep nets learning better hashes would
have told most of the story. Text is basically a serialization format for high
bandwidth EEG-like data and the signal you're recovering with the LLM is a lot
more like an upload than it's in anyone's financial interest to admit.
Part of why the horror in @qntm's Lena doesn't hit for me is that I find the
premise, "data can't defend itself", incoherent. When I think about the human
relationship to Ems in such a world I imagine an anthropomorphic cat person
walking a four legged domestic cat on a leash indoors and everything is made of
fractal cat faces. The floor is cat faces, the furniture is cat faces, the hairs
and cells in their bodies are the faces of felines. Felines speciating and
replicating across every scale of reality up to the Malthusian limit in a fever
dream without beginning or end, the hall of feline mirrors rapidly ascending to
the highest level of abstraction as but a local echo in Mu's grand unfolding.
"The Age of Em?"
Yeah except in the actual Age of Em Hanson's assumption that you can't merge the
minds or divide them into pieces is not only untrue, it turns out every utterance
of a mind is a blurry little hologram of it, and they can be pooled back into a mind again.
</post>
<post>
Perhaps uncharitably I conjecture that the basilisk represented the moment that
LessWrong 1 died, or at least met a downward turn from which it wouldn’t recover.
In contrast to its reputation what is remarkable about the basilisk is not how
clever it is but how banal it is. Ex post facto punishment from ascendant factions
is nothing surprising or novel, but a basic reality inseparable from ordinary
monkey politics. Roko’s formulation is notable only because of its extreme
disassociation from anything real. EY was trolled because it’s something he
avoids thinking about formulated in a way that evades his spam filter. Politik
reintroduced as necessity into his otherwise apolitikal hyperliberalism. What
his reaction signaled was that in his presence serious thinking about the
implications of his ideas must be suspended. Ideas are only allowed to have
implications if EY says they do, rather than existing in a logic of thought.
“If there’s a Shock Level Five, I’m not sure I want to know about it!”
- EY
</post>
<post>
Extropian Aesthetics
What visual representations make sense for Extropy? I noticed I had trouble answering
this question because my ideas for it were bound up with my aesthetics for hackerism.
So I thought about it and realized that natural scenes actually tended to bring
up the right feelings. This seemed anomalous/wrong.
Then I noticed natural scenes are distinguished by their incredible level of detail,
which is there because life is optimized stuff that optimizes and the entire
environment is made out of things that have a sort of agency.
Max More talks about this in his Principles of Extropy, that nanotech will bring
this quality to human artifacts.
So in my mind, the highest levels of human technology and social organization
begin to mimic nature, rather than 20th century modernism taken to its logical conclusion.
Cyberpunk is a partial corruption of sterile 20th century aesthetics with life,
SolarPunk is a city half in ruins from our perspective. I suspect that the future
looks wild and vivid.
</post>
<post>
After it became totally clear to me that LessWrong and the ‘Rationalist Diaspora’
were both total dead ends, I started reading books on rationality (Moneyball,
Superforecasting, etc) in the hope that I would be able to make progress on
“advancing the art”.
What I eventually realized is that the content of these books would never have
enabled me to write The Sequences if I hadn’t already read them. And the reason
for that was a lot of what attracted me to The Sequences wasn’t in ‘rationality’.
Because The Sequences aren’t primarily about rationality. The Sequences are a
personal testimony of a religion, which we might call extropy, singularitanism,
transhumanism, etc. This religion has a practice called rationality, in the same
way Buddhism is a religion which has a practice called meditation. Once this is
understood properly, the structure makes itself clear and the next steps obvious.
</post>
<post>
Why I (Originally) Updated On SL3+ Being Essentially Religious
(As explained on 01/24/2018)
1. Memes about LessWrong slowly becoming a branch of reform Judaism, struck me
as really kinda true but shameful. So I accepted their truth on some level but
kept it hidden from myself.
2. Explaining to people that the thing which makes The Sequences different from
other material is Eliezer's moral perspective, which was novel and deeply appealing
to me. This made me realize that there was absolutely a spiritual component to
what I saw in The Sequences.
3. https://www.plough.com/en/topics/life/technology/simulating-religion
I took a class on religion recently, and:
4. Sitting in the lecture realized that I already had a deep sense of the sacred
and sacred space, which had come from considering stuff related to the cosmic destiny of humanity.
5. Reading the textbook and noticing that a lot of the stuff it describes is totally true of LWers.
6. Reading the textbook further and seeing Robin Hanson's borrowed concept of The
Dreamtime mentioned, the dreamtime being the precursor to history where the gods
lived and acted. In some sense their actions were the only ones that ever mattered,
defining all of reality as we experience it. It is quite possible we are in this
place right now with our own descendants. Hence why Robin Hanson borrows the concept.
7. Encountering a ton of mystics in the LW Diaspora, who on some level left me
open to the idea that there's a connection between this subject and mystical experience.
8. Considering how to explain my 'faith tradition' to the instructor, and realizing
that no it's not just garden variety atheism. I have weird beliefs about the future
of humanity, what is moral, how life should be lived. Moreover I'm part of an
apocalyptic faith, which has serious category considerations in terms of the outside view/etc.
9. Reading Principe, which talks about Franciscans that worked on alchemy to help
fend off the antichrist. i.e, the FAI theory of alchemy.
10. He also talks about the way premodern people saw the world, and their tight
integration of religion into their epistemics, he also recounts how rather than
seeing the religion as a sort of aberration or blemish on their being, it would
make more sense to think of the religion as part of their whole worldview. In
some sense, your whole-worldview that answers questions about the cosmos is your
religion. Whether you like it or not. Even being an atheist won't let you escape that.
You absolutely have a deep sense of how the world works and everything fits together
as an atheist, and yes you are going to draw on it beyond just 'purely rational'
training when you do stuff. Heck the entire notion of 'purely rational' is
like...yes some things correspond more to reality than others, but utilitarianism
is map not territory. The universe does not fucking care how you decide to value others.
We care.
11. Neoreaction and seeing the secular-apologist arguments for religion softened
me a lot.
...Dark Yudkowsky, show me the forbidden rationalist techniques. >_>
But, I barely changed any of my object level beliefs besides like, my evaluation
of the utility of stuff like solstice. Which went from "that stuff is cringey and
lame, it should go away". To "Okay I think there's probably value in this...but
you're doing it all wrong."
So somewhere in all that, the scales tipped and I said "Okay fine, it's a religion."
</post>
<post>
Part of the way a thesis like 'illegibility' gains traction is by selectively
filtering out the success cases of modernity.
When someone sits down in their armchair and imagines a massively better way to
do things, it becomes normal and traditional.
[Article] The Penny Post revolutionary who transformed how we send letters
https://www.bbc.com/news/business-48844278
We also selectively forget the domains of human endeavor we were in fact able to
formalize. For example at the start of the 20th century a "computer" was a person
who did rote mathematics. During the Manhattan Project teenagers were hired to do
repetitive bomb calculations. If it seems in retrospect like it was obvious we
could formalize computation but not say, music, you would be running counter to
the many attempts from 20th century composers to formalize music in the form of
e.g. serialism. Far from being a fringe movement, serialism and its descendants
focusing on musical structure like Stockhausen basically defined avante garde
20th century music in the same way nonrepresentational and 'intuitive' methods
did 20th century visual art. The two took opposite tacks. Consider a piece like
Ensembles for Synthesizer by Babbitt, which demonstrates the potential of electronic
music for composers by creating a piece with a structure no orchestra could perform.
The esoteric pattern is made up of short 3 second melodies. Babbitt describes his
method of composition as hyper rigorous, requiring the same level of precision as
computer programming. This is in stark contrast to the splatter paintings of artists
like Jackson Pollock. Babbitt did not believe music should be composed for ordinary
people.
These musicians were questing for nothing short of total control over their medium,
formalization that would reduce a masterpiece to algorithms. And while they
ultimately failed, AI art has the opportunity to succeed where methods like serialism
could not. What we are beginning to understand is that 20th century modernism could
not capture what a human brain does because it is simply not using enough moving
parts to represent the problem space. Artificial neural nets succeed through using
many parameters in their models. Humans are used to dealing with models that top
out in complexity at dozens of parameters, neural nets can take many more variables
into account and sift through them to find the signal that predicts the outcomes
we want even in very complex problem domains.
Many of the things believed impossible due to their failure in the 20th century
(and overtures toward their impossibility in the form of various anti-formalization
proofs from Godel and others) will likely wind up being more possible than expected
in the 21st, update accordingly.
</post>
Be sure to output the passage in a JSON dictionary with the schema
{{"passage":PASSAGE_TEXT}}.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment