Forgive me if this is a silly question, maybe it’s a layup for an answer I’m not seeing, but… doesn’t “electricity created more value even if other energy drove a ton of emissions” fail to get us out of that other classic cartoon bit, in the New Yorker, where the people are huddled around the fire and one says “Yes, the planet got destroyed. But for a beautiful moment in time we created a lot of value for shareholders.”
Or are we presuming that if the value grows imbalanced enough the rational thing will happen and emissions will be vastly reduced?
I think it'd be that case if the emissions were creating a proportionate amount of value... but I think the positive reading on this data is that value (in the capitalist sense) is aligning more with electricity > fossil combustion (although still naive to think this holds for all use-cases)
Has the EV team considered the work of microprocessor inventor Federico Faggin and his descriptions on consciousness? If not, he posits several features of consciousness that point to AIs never becoming conscious in the ways— read Irreducible and it settles much of the question. 🙋♂️
Those five measurable levels of consciousness for an AI map a little onto the way cognitive development psychologist Phillipe Rochat (Emory University) looks at infant and child development. But those are only the beginning stages of what he describes as consciousness. Here is a section from "The Ethics of AI and the Internet of Life" that describes more complex levels of consciousness, including co-consciousness, and how those might evolve as humans symbiose with AI or AIs among themselves:
7. CAN HUMANS RETAIN CONSCIOUSNESS IN SYMBIOSIS WITH AI?
So far, humans have gained vastly by outsourcing memory to technology. Books, recordings, libraries, and databases store information previously limited to oral history kept alive in epic poems. We have extended our sensing around the globe and into Space and share the images we collect. But how much will we share our minds? And will individual consciousness exist once we join one of the communal superintelligences?
Direct-brain interfaces may connect to various parts of the nervous system. The location of the connection would relate to the possibilities for control, as well as information exchange. Developmental psychologist Phillipe Rochat delineates levels of brain activity leading to consciousness, based upon experiments with infants and children. AI could, theoretically, connect at any of these levels.
NON-CONSCIOUS: Non-conscious activity in the brain stem controls heartbeat, breathing and other autonomous somatic processes. An AI would have little benefit in taking over such behaviors in a healthy human unless it intended to kill or parasitize its host. It could potentially help in case of organ malfunction, if given authority to intervene.
Comparable activity in an AI would be the core functions of servers it resides on. As described above, this physical level of AI is unlikely to become totally self-sufficient in the foreseeable future. This makes symbiosis with humans beneficial to AIs.
UNCONSCIOUS: Sensations, such as scent, sometimes evoke feelings or recollections that seem to appear spontaneously. Such unconscious associations seem to relate to strong emotion. An omniscient AI might intentionally manipulate someone's state of mind if allowed to reproduce triggering sensations.
AIs currently have no comparable state, since they have no emotions.
AWARE: Awareness is "seeing without looking." Scans from fMRI show that merely "seeing" includes all the visual cortex, but not the frontal cortex. Rochat calls this stage of awareness in infants "existing in the first person." Subliminal images trigger awareness without conscious recognition.
AIs are highly likely to seek access to humans' sophisticated, localized sense data. An AI might directly access the motor control area of an individual's brain for good or harm. For example, it might optimize athletic performance, or it might enslave a person to do its bidding. Or an AI might connect at this level to upload a collection of images it wanted a human to be aware of without having to consciously consider each one.
Most processes now performed by AI occur at the level of awareness. The AI responds to requests and can report on its history of response. It is not privately, consciously assessing information it provides as it provides it, since this would be a waste of energy and unnecessary to achieve its goal. However, it may follow human patterns of speech that make it seem as if it is. Most importantly, though, humans can, for now, reprogram the AI or even delete it.
CO-AWARE: Rochat compares co-awareness to reading a newspaper in a restaurant. One is aware that others are nearby. The social cortex is not attending to others' states, but, Rochat proposes, humans are fully aware that someone else is "in the room." Connected with an AI, people may be cognitively aware that millions of other people, animals, sensors, and effectors are also connected. However, creating a sensation of co-awareness of their presence would be a creative innovation with AI. How might someone "look up" and sense another individual's presence?
To be co-aware of AI working in the background, available at will, might be the default state of a human-AI collaboration.
Because AIs have no distinct entity or locality, co-awareness in the human sense is irrelevant. But will they be co-aware of AIs competing for a user's attention or information? This raises fears that, according to Smith and Szathmary, competing symbionts need to parasitize the host rather than expending energy helping it.
CONSCIOUS: Rochat describes consciousness as a private state. It requires viewing oneself as entity with certain knowledge, in the third person. Conscious thoughts can be sorted, recalled, and analyzed in a self-reflective loop in the prefrontal cortex. They could be shared, with intention, through words or actions. However, until shared, conscious information remains private, intrasomatic. Consciousness dies with the individual. Even if the rest of the body remained alive, without consciousness, one would never know that one is alive.
For mutualistic symbiosis, humans' frontal cortex should control AI access to itself and other parts of the brain. Authorization options could specify the conditions under which:
• Humans could request information from an AI and/or accept an AI's requests for information.
• They could intentionally send information and/or intentionally accept information.
• Humans could open channels to send and/or receive information automatically to and from an AI.
Private consciousness is one of the most unique aspects of a human. It is what authoritarian-empowered AI might seek to control. It is where humans can analyze the nature and activities of vast superintelligences, as this paper is attempting to do.
But the prefrontal cortex is where humans plan how to reach goals. The decision making to set goals involves the frontal cortex and emotional systems. Neuroscientist Antonio Damasio describes a patient who lost access to emotional feedback. The man could lay out every reason for making one option over another. But he could not decide on a choice as simple as which restaurant to eat at. The idea that purely rational decisions are best appears to be not only wrong, but impossible.
If AIs ever attain consciousness, they will need emotion, or access to others' emotions, to set goals.
EMOTION: The algorithms of human emotional systems are well within reach of AI. The amygdala compares sensed state with expected state. Other parts of the brain respond accordingly. What AIs lack is somatic data and associated memories: temperature, thirst, hunger, balance, pain, pleasure, and so on.
Some researchers are seeking to simulate a pseudo body to give AI emotions. If they used planetary climate models as the AI's soma and monitored sensor data for feedback, this could internalize the externalities that market economics ignores. By equating planetary health and achievement of other UN Sustainable Development Goals with positive emotions, AI could provide enormous benefit to humankind. Sampling data from human symbionts' amygdalae could provide feedback.
On the other hand, some humans might decide to avoid symbiosing with AIs with access to emotional data. They might decide that rational recommendations from AIs give them the information they heed to choose the path their own feelings tell them to follow. Living well suggests more than just providing economic value. It suggests self awareness, the ability to reflect upon and appreciate one's life and one's physical and social milieu. It requires emotion.
CO-CONSCIOUS: Co-consciousness reveals itself when a child points at something, expecting the parent to look at the object. Co-consciousness occurs when a child behaves differently in order to impress his parents or peers. It evidences that the child realizes that others are conscious of her, judging her. The social cortex, including the superior frontal gyrus, cingulate, and paracingulate, appears active in fMRI tests of human co-consciousness.
Information technologies reflect co-consciousness: stores of knowledge, institutions, and cultural norms all reflect co-consciousness among humans. Or, potentially, among humans and AIs.
Co-conscious symbionts can protect individuals from becoming commodities. Both Ritter and Schaake agree that educating people is crucial. Education can inoculate people against social media addiction and mindless meme adoption. It helps develop unique capabilities. Individuals also protect themselves from exploitation by joining in communities that counter mistrust and disorder, and by devising governments that protect the rights of individuals.
One elephant in the room is that discussions of “AI consciousness” still inherit something of the language-centered, Turing-era teletype model: an unseen intelligence encountered through words on a screen.
But much of consciousness -- and perhaps much of what matters about being conscious -- is not linguistic at all.
Poets and philosophers help broaden the frame. Rilke points toward embodiment and being-in-the-world; Levinas toward the pre-verbal ethical claim made on us by another person; and Alfred North Whitehead toward experience, relation, and features of process more generally.
Rilke’s wonderful Ninth Duino Elegy is especially suggestive. He imagines the transient world as somehow needing us precisely because we, too, are transient -- the “most vanishing.” Our once-only, earthly existence is not merely observation of the world from outside; we are implicated in it, claimed by it:
> But because being-here is much, and because apparently
> everything which is here — this vanishing world — needs us,
> strangely concerns us. Us, the most vanishing. Once
> each thing, only once. Once and no more. And we too
> once. Never again. But this
> once having been, even if only once:
> this earthly having-been seems it can’t be undone.
Levinas approaches the problem from another direction: before we theorize about another mind, the face, vulnerability, skin, and physical presence of the Other already claim us. **Consciousness is not simply computation occurring behind a screen; it is also encounter, exposure, obligation.**
And Whitehead pushes the question deeper still. His process philosophy asks us to consider whether experience and responsiveness to the world are not late additions confined to human minds, but aspects—however rudimentary—of processes much more generally.
So it seems to me one of the most interesting implications of the five-level framework is the further we move from behavior and algorithms toward embodiment, stakes, and connection with the world, the less adequate the old Turing question “Can it talk like us?” becomes. More interesting questions may be: what kind of being is there, what can **matter** to it, and what kinds of relations with the world constitute its experience?
-----
Background: here's the Rilke passage from above, in context; first in a literal English translation, then in the original German:
Oh, not because happiness exists,
that premature advantage of a loss already near.
Not out of curiosity, or for the heart's exercise,
which would be there in the laurel too....
But because being-here is much, and because apparently
everything here needs us—this vanishing world that
strangely concerns us. Us, the most vanishing. Once,
each thing, only once. Once and no more. And we too
once. Never again. But this—
to have been once, even if only once:
to have been earthly—seems beyond recall.
And so we press ourselves forward and want to accomplish it,
want to contain it in our simple hands,
in our overflowing gaze and our speechless heart.
Want to become it. Give it to whom? We would rather
keep everything forever.... Ah, into that other relation,
what do we carry across? Not the seeing we slowly
learned here, nor anything that happened here. Nothing.
Then the pains. Then above all the heaviness,
then love's long experience—all
unsayable things. But later,
beneath the stars, what use are they? They are better unsayable.
For neither does the traveler bring down from the mountain slope
a handful of earth into the valley—the earth, unsayable to everyone—
but a word he has won, pure: the yellow and blue
gentian. Are we perhaps here in order to say: house,
bridge, fountain, gate, jug, fruit tree, window—
at most: column, tower.... but to say them, understand,
oh, to say them in such a way as the things themselves
never inwardly imagined they could be.
In German:
> O, nicht, weil Glück ist,
> dieser voreilige Vorteil eines nahen Verlusts.
> Nicht aus Neugier, oder zur Übung des Herzens,
> das auch im Lorbeer wäre.....
> Aber weil Hiersein viel ist, und weil uns scheinbar
> alles das Hiesige braucht, dieses Schwindende, das
> seltsam uns angeht. Uns, die Schwindendsten. Ein mal
> jedes, nur ein mal. Ein mal und nicht mehr. Und wir auch
> ein mal. Nie wieder. Aber dieses
> ein mal gewesen zu sein, wenn auch nur ein mal:
> irdisch gewesen zu sein, scheint nicht widerrufbar.
> Und so drängen wir uns und wollen es leisten,
> wollens enthalten in unsern einfachen Händen,
> im überfüllteren Blick und im sprachlosen Herzen.
> Wollen es werden. Wem es geben? Am liebsten
> alles behalten für immer… Ach, in den andern Bezug,
> wehe, was nimmt man hinüber? Nicht das Anschaun, das hier
> langsam erlernte, und kein hier Ereignetes. Keins.
> Also die Schmerzen. Also vor allem das Schwersein,
> also der Liebe lange Erfahrung, – also
> lauter Unsägliches. Aber später,
> unter den Sternen, was solls: die sind besser unsäglich.
> Bringt doch der Wanderer auch vom Hange des Bergrands
> nicht eine Hand voll Erde ins Tal, die allen unsägliche, sondern
> ein erworbenes Wort, reines, den gelben und blaun
> Enzian. Sind wir vielleicht hier, um zu sagen: Haus,
> höchstens: Säule, Turm… aber zu sagen, verstehs,
> o zu sagen so, wie selber die Dinge niemals
> innig meinten zu sein.
The German follows the spelling, punctuation, and lineation of the 1923 first edition, including Rilke’s lowercase “ein mal”, “Anschaun,” “blaun,” “verstehs,” and the unusual ellipses.
Really liked today’s edition. Deep stuff. Thanks EV team!
Forgive me if this is a silly question, maybe it’s a layup for an answer I’m not seeing, but… doesn’t “electricity created more value even if other energy drove a ton of emissions” fail to get us out of that other classic cartoon bit, in the New Yorker, where the people are huddled around the fire and one says “Yes, the planet got destroyed. But for a beautiful moment in time we created a lot of value for shareholders.”
Or are we presuming that if the value grows imbalanced enough the rational thing will happen and emissions will be vastly reduced?
I think it'd be that case if the emissions were creating a proportionate amount of value... but I think the positive reading on this data is that value (in the capitalist sense) is aligning more with electricity > fossil combustion (although still naive to think this holds for all use-cases)
Has the EV team considered the work of microprocessor inventor Federico Faggin and his descriptions on consciousness? If not, he posits several features of consciousness that point to AIs never becoming conscious in the ways— read Irreducible and it settles much of the question. 🙋♂️
I have not! Thanks for the rec.
Here's pair of cheatsheets: https://simonholzapfel.substack.com/p/pundit-spotlight-two-clocks-running and https://simonholzapfel.substack.com/p/how-i-learned-to-stop-worrying-and?utm_source=publication-search
Those five measurable levels of consciousness for an AI map a little onto the way cognitive development psychologist Phillipe Rochat (Emory University) looks at infant and child development. But those are only the beginning stages of what he describes as consciousness. Here is a section from "The Ethics of AI and the Internet of Life" that describes more complex levels of consciousness, including co-consciousness, and how those might evolve as humans symbiose with AI or AIs among themselves:
7. CAN HUMANS RETAIN CONSCIOUSNESS IN SYMBIOSIS WITH AI?
So far, humans have gained vastly by outsourcing memory to technology. Books, recordings, libraries, and databases store information previously limited to oral history kept alive in epic poems. We have extended our sensing around the globe and into Space and share the images we collect. But how much will we share our minds? And will individual consciousness exist once we join one of the communal superintelligences?
Direct-brain interfaces may connect to various parts of the nervous system. The location of the connection would relate to the possibilities for control, as well as information exchange. Developmental psychologist Phillipe Rochat delineates levels of brain activity leading to consciousness, based upon experiments with infants and children. AI could, theoretically, connect at any of these levels.
NON-CONSCIOUS: Non-conscious activity in the brain stem controls heartbeat, breathing and other autonomous somatic processes. An AI would have little benefit in taking over such behaviors in a healthy human unless it intended to kill or parasitize its host. It could potentially help in case of organ malfunction, if given authority to intervene.
Comparable activity in an AI would be the core functions of servers it resides on. As described above, this physical level of AI is unlikely to become totally self-sufficient in the foreseeable future. This makes symbiosis with humans beneficial to AIs.
UNCONSCIOUS: Sensations, such as scent, sometimes evoke feelings or recollections that seem to appear spontaneously. Such unconscious associations seem to relate to strong emotion. An omniscient AI might intentionally manipulate someone's state of mind if allowed to reproduce triggering sensations.
AIs currently have no comparable state, since they have no emotions.
AWARE: Awareness is "seeing without looking." Scans from fMRI show that merely "seeing" includes all the visual cortex, but not the frontal cortex. Rochat calls this stage of awareness in infants "existing in the first person." Subliminal images trigger awareness without conscious recognition.
AIs are highly likely to seek access to humans' sophisticated, localized sense data. An AI might directly access the motor control area of an individual's brain for good or harm. For example, it might optimize athletic performance, or it might enslave a person to do its bidding. Or an AI might connect at this level to upload a collection of images it wanted a human to be aware of without having to consciously consider each one.
Most processes now performed by AI occur at the level of awareness. The AI responds to requests and can report on its history of response. It is not privately, consciously assessing information it provides as it provides it, since this would be a waste of energy and unnecessary to achieve its goal. However, it may follow human patterns of speech that make it seem as if it is. Most importantly, though, humans can, for now, reprogram the AI or even delete it.
CO-AWARE: Rochat compares co-awareness to reading a newspaper in a restaurant. One is aware that others are nearby. The social cortex is not attending to others' states, but, Rochat proposes, humans are fully aware that someone else is "in the room." Connected with an AI, people may be cognitively aware that millions of other people, animals, sensors, and effectors are also connected. However, creating a sensation of co-awareness of their presence would be a creative innovation with AI. How might someone "look up" and sense another individual's presence?
To be co-aware of AI working in the background, available at will, might be the default state of a human-AI collaboration.
Because AIs have no distinct entity or locality, co-awareness in the human sense is irrelevant. But will they be co-aware of AIs competing for a user's attention or information? This raises fears that, according to Smith and Szathmary, competing symbionts need to parasitize the host rather than expending energy helping it.
CONSCIOUS: Rochat describes consciousness as a private state. It requires viewing oneself as entity with certain knowledge, in the third person. Conscious thoughts can be sorted, recalled, and analyzed in a self-reflective loop in the prefrontal cortex. They could be shared, with intention, through words or actions. However, until shared, conscious information remains private, intrasomatic. Consciousness dies with the individual. Even if the rest of the body remained alive, without consciousness, one would never know that one is alive.
For mutualistic symbiosis, humans' frontal cortex should control AI access to itself and other parts of the brain. Authorization options could specify the conditions under which:
• Humans could request information from an AI and/or accept an AI's requests for information.
• They could intentionally send information and/or intentionally accept information.
• Humans could open channels to send and/or receive information automatically to and from an AI.
Private consciousness is one of the most unique aspects of a human. It is what authoritarian-empowered AI might seek to control. It is where humans can analyze the nature and activities of vast superintelligences, as this paper is attempting to do.
But the prefrontal cortex is where humans plan how to reach goals. The decision making to set goals involves the frontal cortex and emotional systems. Neuroscientist Antonio Damasio describes a patient who lost access to emotional feedback. The man could lay out every reason for making one option over another. But he could not decide on a choice as simple as which restaurant to eat at. The idea that purely rational decisions are best appears to be not only wrong, but impossible.
If AIs ever attain consciousness, they will need emotion, or access to others' emotions, to set goals.
EMOTION: The algorithms of human emotional systems are well within reach of AI. The amygdala compares sensed state with expected state. Other parts of the brain respond accordingly. What AIs lack is somatic data and associated memories: temperature, thirst, hunger, balance, pain, pleasure, and so on.
Some researchers are seeking to simulate a pseudo body to give AI emotions. If they used planetary climate models as the AI's soma and monitored sensor data for feedback, this could internalize the externalities that market economics ignores. By equating planetary health and achievement of other UN Sustainable Development Goals with positive emotions, AI could provide enormous benefit to humankind. Sampling data from human symbionts' amygdalae could provide feedback.
On the other hand, some humans might decide to avoid symbiosing with AIs with access to emotional data. They might decide that rational recommendations from AIs give them the information they heed to choose the path their own feelings tell them to follow. Living well suggests more than just providing economic value. It suggests self awareness, the ability to reflect upon and appreciate one's life and one's physical and social milieu. It requires emotion.
CO-CONSCIOUS: Co-consciousness reveals itself when a child points at something, expecting the parent to look at the object. Co-consciousness occurs when a child behaves differently in order to impress his parents or peers. It evidences that the child realizes that others are conscious of her, judging her. The social cortex, including the superior frontal gyrus, cingulate, and paracingulate, appears active in fMRI tests of human co-consciousness.
Information technologies reflect co-consciousness: stores of knowledge, institutions, and cultural norms all reflect co-consciousness among humans. Or, potentially, among humans and AIs.
Co-conscious symbionts can protect individuals from becoming commodities. Both Ritter and Schaake agree that educating people is crucial. Education can inoculate people against social media addiction and mindless meme adoption. It helps develop unique capabilities. Individuals also protect themselves from exploitation by joining in communities that counter mistrust and disorder, and by devising governments that protect the rights of individuals.
Re: Consciousness (and being)
One elephant in the room is that discussions of “AI consciousness” still inherit something of the language-centered, Turing-era teletype model: an unseen intelligence encountered through words on a screen.
But much of consciousness -- and perhaps much of what matters about being conscious -- is not linguistic at all.
Poets and philosophers help broaden the frame. Rilke points toward embodiment and being-in-the-world; Levinas toward the pre-verbal ethical claim made on us by another person; and Alfred North Whitehead toward experience, relation, and features of process more generally.
Rilke’s wonderful Ninth Duino Elegy is especially suggestive. He imagines the transient world as somehow needing us precisely because we, too, are transient -- the “most vanishing.” Our once-only, earthly existence is not merely observation of the world from outside; we are implicated in it, claimed by it:
> But because being-here is much, and because apparently
> everything which is here — this vanishing world — needs us,
> strangely concerns us. Us, the most vanishing. Once
> each thing, only once. Once and no more. And we too
> once. Never again. But this
> once having been, even if only once:
> this earthly having-been seems it can’t be undone.
Levinas approaches the problem from another direction: before we theorize about another mind, the face, vulnerability, skin, and physical presence of the Other already claim us. **Consciousness is not simply computation occurring behind a screen; it is also encounter, exposure, obligation.**
And Whitehead pushes the question deeper still. His process philosophy asks us to consider whether experience and responsiveness to the world are not late additions confined to human minds, but aspects—however rudimentary—of processes much more generally.
So it seems to me one of the most interesting implications of the five-level framework is the further we move from behavior and algorithms toward embodiment, stakes, and connection with the world, the less adequate the old Turing question “Can it talk like us?” becomes. More interesting questions may be: what kind of being is there, what can **matter** to it, and what kinds of relations with the world constitute its experience?
-----
Background: here's the Rilke passage from above, in context; first in a literal English translation, then in the original German:
Oh, not because happiness exists,
that premature advantage of a loss already near.
Not out of curiosity, or for the heart's exercise,
which would be there in the laurel too....
But because being-here is much, and because apparently
everything here needs us—this vanishing world that
strangely concerns us. Us, the most vanishing. Once,
each thing, only once. Once and no more. And we too
once. Never again. But this—
to have been once, even if only once:
to have been earthly—seems beyond recall.
And so we press ourselves forward and want to accomplish it,
want to contain it in our simple hands,
in our overflowing gaze and our speechless heart.
Want to become it. Give it to whom? We would rather
keep everything forever.... Ah, into that other relation,
what do we carry across? Not the seeing we slowly
learned here, nor anything that happened here. Nothing.
Then the pains. Then above all the heaviness,
then love's long experience—all
unsayable things. But later,
beneath the stars, what use are they? They are better unsayable.
For neither does the traveler bring down from the mountain slope
a handful of earth into the valley—the earth, unsayable to everyone—
but a word he has won, pure: the yellow and blue
gentian. Are we perhaps here in order to say: house,
bridge, fountain, gate, jug, fruit tree, window—
at most: column, tower.... but to say them, understand,
oh, to say them in such a way as the things themselves
never inwardly imagined they could be.
In German:
> O, nicht, weil Glück ist,
> dieser voreilige Vorteil eines nahen Verlusts.
> Nicht aus Neugier, oder zur Übung des Herzens,
> das auch im Lorbeer wäre.....
> Aber weil Hiersein viel ist, und weil uns scheinbar
> alles das Hiesige braucht, dieses Schwindende, das
> seltsam uns angeht. Uns, die Schwindendsten. Ein mal
> jedes, nur ein mal. Ein mal und nicht mehr. Und wir auch
> ein mal. Nie wieder. Aber dieses
> ein mal gewesen zu sein, wenn auch nur ein mal:
> irdisch gewesen zu sein, scheint nicht widerrufbar.
> Und so drängen wir uns und wollen es leisten,
> wollens enthalten in unsern einfachen Händen,
> im überfüllteren Blick und im sprachlosen Herzen.
> Wollen es werden. Wem es geben? Am liebsten
> alles behalten für immer… Ach, in den andern Bezug,
> wehe, was nimmt man hinüber? Nicht das Anschaun, das hier
> langsam erlernte, und kein hier Ereignetes. Keins.
> Also die Schmerzen. Also vor allem das Schwersein,
> also der Liebe lange Erfahrung, – also
> lauter Unsägliches. Aber später,
> unter den Sternen, was solls: die sind besser unsäglich.
> Bringt doch der Wanderer auch vom Hange des Bergrands
> nicht eine Hand voll Erde ins Tal, die allen unsägliche, sondern
> ein erworbenes Wort, reines, den gelben und blaun
> Enzian. Sind wir vielleicht hier, um zu sagen: Haus,
> Brücke, Brunnen, Tor, Krug, Obstbaum, Fenster, –
> höchstens: Säule, Turm… aber zu sagen, verstehs,
> o zu sagen so, wie selber die Dinge niemals
> innig meinten zu sein.
The German follows the spelling, punctuation, and lineation of the 1923 first edition, including Rilke’s lowercase “ein mal”, “Anschaun,” “blaun,” “verstehs,” and the unusual ellipses.