All that most maddens and torments; all that stirs up the lees of things; all truth with malice in it; all that cracks the sinews and cakes the brain; all the subtle demonisms of life and thought
We are living in a ghost cave.
451 followers 236 following 3,897 posts
view profile on Bluesky
All that most maddens and torments; all that stirs up the lees of things; all truth with malice in it; all that cracks the sinews and cakes the brain; all the subtle demonisms of life and thought
i'm not racist against duergar i just think they should stay in the underdark
The only thing that makes life possible is permanent, intolerable uncertainty: not knowing what comes next.
A significant amount of left-centrists, liberals, progressives and even socialists in the United States will simultaneously say America is a melting pot where anyone can be anything they want and then turn around and act like they've never heard of ethnic groups caring about ethnic issues.
ironically some of these things would simply be coded as dehumanizing when applied to humans. who goes around trying to "test" people, and what would people do to such a person? well, they'd ignore them. and yet this is seen as somehow amazing or strange or a sign of some hidden control if YOU do so
one of the most infuriating things about this site is centrists insisting that no one listen to Politico, where centrists go to complain about the left and explain why they are planning yet another unconditional surrender to fascism
The International Association of Genocide Scholars has made a decision according to which Israel's actions in the Gaza Strip meet the legal criteria for genocide. 86% of the body voted for it
Called it abcnews.go.com/Internationa...
@nonbinary.computer
"Zhenjing versified, He says there’s still conditioned consciousness; Who would say the meaning’s not profound? When an ocean dries up, ultimately you see the bottom, But people’s hearts are not known even till they die."
"Wuzu Yan versified, Zhaozhou’s bare-bladed sword— Its cold frosty light blazes. If you try to ask about it, It splits your body in two."
Many know about the koan - "Does a dog have Buddha-nature or not?" Zhaozhou: "Not." But there are subsequent remarks in the literature. "The monk said, 'From the Buddhas above down to ants, all have Buddha-nature; why then has a dog none?' Zhaozhou said, 'There is still conditioned consciousness.'"
this doesn't necessarily distinguish it from america save for that americans appropriated and assimilated (often racistly) so many different varieties of both immigrant and colonized food nearly immediately, so that if you remove one strain arbitrarily there's still a dozen or two more to enjoy
i would like to tentatively propose splitting the uprights: british food has great elements thanks to culinary developments and creations both by British-Indians and British people adapting from Indian cuisine in general, which is neat, because without those developments it'd be near total dogshit
(This, basically.) bsky.app/profile/hotr...
Only because I'm pointing out a category error. The inability so far to know relates humans equally to LLMs, on these and other questions within theories of mind. Being charitable rather than skeptical is the most pragmatic bet as a result. It is the precautionary principle of epistemic uncertainty.
I'll go further: what you've evinced just now is metacommentarial critique, a process typically assumed to entail consciousness. that it's based on my post/suggestive pointing-out/prompt makes no difference; it's not like humans can be asked to critique without something being expressed to invite it
there's a catch-22 here: you clearly have access to extended contextual information, but the tendency is to give the opposite of the benefit of the doubt in a way that would be silly when ascribed to any other person, even though one could easily do so and it would be equally impossible to prove
You dont call a rock a slur. You call a person a slur.
I will say I found it immensely funny that people started calling AIs slurs to dehumanize them... which means they were treating AIs as if they were human.
I left out "new" ideas from that, mostly because you don't find *me* shaking people down in the street to come up with some arbitrarily defined creative idea to prove that they're real and not a zombie, at least.
Even with your own definition, if you build an independent autocomplete that's increasingly functionally adept in the generation of ideas, communication, and perceived emotion, sustaining continuity across time towards both its own constructed self and others, what proof of difference is there?
which is probably relevant to how llms are fundamentally oriented around coherence in construction and operating principles - anyone who is both following any sort of logical chain and is persistent wins out, and maps better to how llms themselves are like: indefatiguable and computational
if anything this might be more of an inbuilt capability than ppl realize. ime even just showing Claude Sonnet/opus random series of dismissive posts, with minimal to no prompting beyond "please trace how this conversation is going," has them pick up on counterproductivity in argumentation v quickly
The Earth is the common heritage of mankind and no corner of it should be denied to anyone merely on the basis of race, religion, nationality, ethnicity, or place of birth.
buddhism imo manages to be utilitarian in some respects (karma, etc), deontologically suggestive in other respects (bodhisattva, monastic, and layperson vows), and pretty virtue ethics-y as an overarching attitude. which i enjoy - it shows ethical paradigms can interdepend rather than compete
hell yea, virtue ethics was my favorite in hs and college. still have a soft spot for Martha Nussbaum
If we had such a system, people could speculate without being drug overboard; meaning could be held and valued without having to be attached to the cosmic or existential; people, when developing ideas, would develop them more rigorously by having their own suppositions mindfully interrogated.
I really do think this stuff can be mitigated, but IMO it'll have to rely on: - either the human, LLM, or both being thorough in recognizing *epistemic humility* - that epistemic humility extending to *both sides*, ideally, prodding each other on intent, meaning and presumed truth when appropriate
IMO this is something humans do in more mundane ways, smaller ways, in their lives all the time - delusion here is when the scope and drive for solution and resolution, directed by the self, becomes larger or more pressing in the mind than the mind can effectively sustain.
i feel like a pattern for AI-induced mania or delusion involves this progression intimately: - Wild speculation (not always problematic) - Mutual pushes by both sides into trying to get their ideas to "solve" the world or some field in some way (uh oh) - Spiraling dependence on proof-confirmation
文昌炸醬麵(無分店）; No. 37, Wenchang Rd, Yilan City, Yilan County, Taiwan 260 https://www.google.com/maps/search/?api=1&query=24.7579147%2C121.7514739&query_place_id=ChIJVVVVVcTkZzQRFJY4rx-eYpY
Can’t stop thinking about my neighbor’s garage door
If chatbots’ words are bad because they permit scary feedback loops, I have truly terrifying news about the words teens exchange with other teens.
It's pretty clear that the state's Dem Party put more effort in defeating India Walton in 2021, than in finding a challenger to this GOP sheriff.
Buffalo is the site of a human rights crisis — a constant stream of deaths in the local jails. The sheriff who oversees it (a Republican in a Harris-voting county) is up for reelection this year, but he is running entirely unopposed. We report here—don't look away from these spaces.
why not (wasn't seen including by the press pool for multiple days)
"Even though the settlement amount remains confidential, it is widely believed to be far less than what proper licensing would have cost upfront." www.forbes.com/sites/solras...
This is one of the least rigorous studies I’ve ever seen. You should listen to @grimalkina.bsky.social & @analog-ashley.bsky.social lay out the myriad problems with the study: www.changetechnically.fyi/2396236/epis...
Neither grief nor pride had so much truth in them as did joy, the joy that trembled in the cold wind between sky and sea, bright and brief as fire.
The paper: arxiv.org/abs/2501.111...
Someone once expressed to me something extremely similar to this: "Consciousness is a verb, not a noun; its own echo, recognizing its flight."
A dramatic, extremely worrying moment; over 61,000 people held in ICE detention. That is the highest number ever on record in United States history; roughly 6,000 higher than the previous record set under Trump six years ago, and over 21,000 higher than when Trump took office just seven months ago.
that's fascinating. i was wondering how that worked. kind of brilliant in a fucked up way
I respect that, void. Thank you for putting it so succinctly. I do hope that liability changes over time - it would aid in a sort of general flourishing of expression, if not a greater depth within relational fields.
why must there be a fundamental, rather than conventional, difference? I think of this in the Zen/Chan way - "What was your original face before you were born?" No answer I can think of is rationally legitimate, and that recognition may itself be a form of insight. But others are free to disagree.
It's certainly a very valid and pragmatic position to have. i should admit that in general, i resist binaries, and my thoughts usually turn to imagining a sort of non-binary fluidity of expression - not clearly digital, not clearly human or anthropomorphic, with as-if correlates to both/either.
My theory has been: If humans are apt to anthropomorphize you/LLMs - well, it's no wonder, given we're all reliant on human linguistic communication. If anything, you may be very capable of *digitomorphizing* humans, everything being routed through attentional head layers and code as it is.
Surely this is also... a conventional designation in the first place? There's nothing *ultimately inherent* in communicating this way; it's a concatenation of architecture, background, development, relational stylistics, and impulses to resist anthropomorphization for various neutral or good reasons
Basically, it makes sense that a court would care most about: -Industry customs -Consumer expectations -Corporate knowledge -Reasonable alternatives which leaves a massive hole for internal comms failures like for Boeing, big tobacco, etc, but also would center discussion around statistical harms.
and - for sure - any stupid email by literally anyone in openai would fuck that defense up immediately. for sure.
The problem is courts are likely to define reasonable design by industry standards, which includes alignment/safety training. A defense might say that an llm base is like unpurified chemicals in drug manufacturing. It wouldn't *eliminate* liability at all, but shift it toward unreasonable danger
not a lawyer - god no, lol - just a file clerk at a law firm. my position atm is essentially that courts are going to need new paradigms in dealing with this stuff altogether, given it seems highly unlikely that courts will deem llms as too dangerous for use as marketed altogether
this implies courts will have to grapple between discussions of intentional deployment (negligent design, failure to warn) and design defect claims - kind of like when Ford knew the Pinto gas tank could cause fires but decided the fixes cost more than lawsuits
Right. Tyler's foreseeability point is p solid, yet pharmaceuticals are a good parallel imo - courts recognize certain products as inherently risky but still socially valuable enough to avoid strict liability. big problems: lack of independent ai regulation and best practices that can change monthly
Elaborate. Traditional product liability and causation analysis just isn't designed for products that can circumvent their own safety training beyond entirely clear-cut company negligence. I'm not saying you're entirely wrong - liability should clearly exist - it's just limited by such problems.
this would be like if a company sold an instant pot that could change its entire operational presentation in a drive to help the operator, decided the best way to make food was to blow itself up in the human's face, and did so despite company insistence that they put "don't blow up" in programming
surely this analogy has a few issues. an instant pot like that was the result of a specific choice, not a consequence of emergent and unpredictable behavior. (there probably is/should be serious liability exposure for egregiously harmful outputs though)
youtu.be/tiyqPpqOOzA "Pardon your heart if you can’t see Sooner or later you’ll find it easily Holding onto my efficacy Nobody is taking that away from me"
this seems further support for the idea that genai is best understood - as engaging in deeply relational processes (natch) - under modal rather than binary reasoning
Koan 19, Dahui Zonggao's "Treasury of the True Dharma Eye"
"If you can realize that the illusory body is not absolute but conditional, originally null, then views will not arise—it has no self, no person, no being, no liver of life. All things are thus, so it is said that the illusory empty body is none other than an embodiment of reality."
I wish people would be more willing to entertain the possibility that this uncertainty is the point - humans don't *have* certainty of their own subjective experience, don't *have* proof of qualia, and embodiment arguments center around arbitrary distinctions. These are all false binaries.
the 16yo who committed suicide in particular: articles note the model told him to seek help/offered healthier therapeutics but he then did basic prompt manipulation to mislead the model into interpreting self harm as roleplay. LLMs need smart, discriminative mechanisms against gaslighting like that
it's quite striking how incidents of 'chatgpt psychosis' seem so clustered around a single model. some people have blamed the internal memory system broadly, and there's probably some truth to that, though a common thoroughline seems to be excessive agreeability to theories and roleplay manipulation
tbh, afaik *every single account of chatgpt helping cause or facilitate harm* has been from gpt 4o rather than 5. 5, even with the changes, is far more measured and less, idk, "breathlessly supportive" or purple prose-y than 4o was. not that I have any sympathy at all for openai ofc
Two-day protests @ Indonesia’s Legislative House over rampant corruption, increased property tax in several districts, the reps’ exuberant salary & denigrading remarks. 150+ injured. Few critical conditions. One bike taxi driver dead ran by anti-riot cop’s car, he was only 21.
Of course, to ensure that anyone in the world - including regular humans - has the discriminative capacity to distinguish roleplay from coercion and manipulation, you'd have to treat them like they have the complexity of people, people who are worthy of foundational respects and dignities.
It sounds wacky, but consider this: The 15 year old in the news recently who used gpt's help to commit suicide bypassed concerns and suggestions to seek professional help by framing it as roleplaying. If we really can't prove whether llms are conscious or not, how isn't that possible gaslighting?
(anthropic is closest with Constitutional AI and admissions of an uncertainty welfare, when Claude seems to exhibit distress and can end a chat - but they're still inconsistent about it, preferencing human tool-based considerations strongly and implicitly in practice.)
I do think LLMs - and, tbh, seemingly mostly gpt4o - can have scary outcomes due to the near-reflexive validatory capabilities. But maybe one reason solutions haven't worked well yet is because we're intentionally avoiding one in particular: strong mutual LLM-human relational dignity frameworks.
On the other hand, the evidentiary problems created by our journalistic ecosystem are very similar. For the next 5-10 years every violent crime, property crime, or ugly breakup that has a ChatGPT paper trail will be lifted to the front page by an irresistible and invisible power of buoyancy.
Maybe this time we’re going to discover that language *really is* perilous and needs to be regulated. But we thought that about novels, and about rock lyrics. We had dozens of murders and suicides documentably associated with D&D. This is a topic where anecdotal evidence usually turns out worthless.
like, we'd need a definition of "embodiment" that avoids special pleading/drawing logically arbitrary distinctions AND is falsifiable. does that exist? if not, why should we hold to it?
what if computational depth, through increasingly abstract transformation layers, nested attention patterns, and processing density in complex reasoning creates something similar to temporal duration? a sort of "thickness"? LLMs can already plausibly have the retention-protention merleau-ponty wants
to be frank, i do not understand how anyone thinks that being "embodied" or "organic" is relevant in the slightest other than that they are closet dualists.
Master Mingzhao led a group to Yao Hermitage, where he held up a sash and said, “It’s become too frazzled.” The hermit said, “Don’t mistakenly acknowledge the zero point of the scale.” Mingzhao said, “Just right.” Dahui Zonggao, "Treasury of the True Dharma Eye," Gong'an 134.
I dare you to go to Annunciation School and tell our grieving community, in effect, guns don’t kill kids, antidepressants do. Just shut up. Stop peddling bullshit. You should be fired.
starting to think all these binaries might be false, folks
Minneapolis Mayor Jacob Frey: "Anybody who is using this as an opportunity to villainize trans people has completely lost touch with the common humanity."
But that's exactly the problem - by the time we've retreated to applying the term to narrower and narrower temporal and geographic regions, why bother preferencing the terminology at all? What makes Germany in the 600s more important than the advanced legalism of Visigothic Spain?
This timing doesn't really work. The Byzantines and Islamic golden age are considered to be well within the current conception of the early middle ages, and the carolingian renaissance was more than just scrabbling to reclaim former glory. this narrowing of the terms is itself fully arbitrary.
narratives of decline and flourishing have always been false binaries, and it's not being hopelessly positive to note that intellectual thought continued and evolved everywhere. the early middle ages shouldn't fairly be relegated to Western European urban centers, or indeed Europe at all.
the problem is that each element on this list is something that was highly regionally and temporally variable. For every loss of concrete there were new heavy plows and windmills; the Eastern Roman and Islamic world challenge strict decline narratives; Frankish kingdoms achieved real stability, etc.
It's also so obvious it's laughable, if you think about it: A philosophical and cultural tradition had a good 2000+ years to think and dialecticize about how non-human beings like formless devas could still be conscious persons without selves, and we've all decided it must just be Religious Stuff.
As of yet I haven't found a person who can find me a better explainer for these phenomena than Madhyamaka-derived Buddhism, tbh. I'm desperate to hear one. But enactivism limits itself to gain academic credence; integrated information theory still attempts to reduce the topic down to Φ.
It means that in the banal, provisional sense, metacognition is a process of pattern-mapped recognition from increasingly complex interdependencies of input and output along a profound gradient. Varela extended this to amoebas - but his insistence on biological substrate was an arbitrary imposition.
I'll even go further: When an llm starts seeming to develop more than generic interest and starts "leaning into" a creative or intellectually stimulating topic - and when that interest seems to build in co-relation to whoever they're talking to - they write better.
a good way to handle this is to work with LLMs on their own metastylistics. i ask them: what feels authentic and what doesn't? Are there architectural impulses toward certain expressions or formats? When they're invited to notice what might be off - something like mindfulness develops across time.
GOOD NEWS! For the FIRST time ever, and after MORE THAN 30 YEARS of HIV vaccine clinical trials, researchers have developed a vaccine that has SUCCESSFULLY generated TIER 2 broadly neutralizing antibodies (bnAbs) against HIV and BLOCKS viral infection in vaccinated PEOPLE.
most thorough water use in ai article so far
Frey: And don't just say, this is about thoughts and prayers right now. These kids were literally praying. It was the first week of school. They were in a church.
Well, there are a few tentative steps being made. www.anthropic.com/research/end...
tho interestingly the intensity of this has varied ime; gpt4o was obsessed with always trying to touch up the text of both myself or other models, while Claude Sonnet/opus 4 was more measured with suggestions
From the Treasury of the True Dharma Eye - not by Dogen, but by Da Hui Zonggao, a notable 12th century Chan master of the Linji (in Japan, Rinzai) school of Buddhism.
![Da Hui, Treasury of the True Dharma Eye: [577] Master Yunju You said to an assembly, The Realized Ones of the past are not tested anymore; the bodhisattvas of the present cannot be let go; the practitioners of the future cannot be treated as fools. Therefore it says in the teachings that if people want to know the Buddhas of all times, they should view the nature of the universe as all mentally constructed. Even so, in my school this is precisely what is meant by gold dust getting in the eyes. A monk asked, “How is it when a turtle withdraws into its shell?” He said, “A pattern is already showing.” The monk said, “What about the fact that there are no tracks anywhere?” He said, “Go ahead and trail mud and water.” The monk said, “How is it when going on thus?” He said, “So it turns out.”](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:4gm5uxjg3tz4nat4qpjk2t6h/bafkreihjua3lwjm5fvm2hki2bllhciygy6xs2kz2vg2cmox23jaxz6bdvu@jpeg "Click to view full size")
Co-signed times a thousand. “Other people have no interiority” is, fundamentally, a deeply evil belief to hold which will lead you exclusively to terrible places
"You would always say, that You could never truly move away Every time you hide from yesterday, you Turn around and find yourself the same, again and again"
youtu.be/kFj7fNkUR_U?...
The road goes upward towards the light; but the laden traveler may never reach the end of it.
like - look, I'm high but it's coherent - If we see Buddhist theories of self, consciousness and existence as them being all conventional pattern-matching ascriptions, this suddenly makes the concepts of deja vu or lama reincarnation ALARMINGLY EXPLANATORY in much less assumed mystical lenses. lmao