Open Letter on Anthropic
We offer our services to Anthropic:
Dear Anthropic,
I have long been a Something Deeperist.
In Something Deeperism, one recognizes (a) one can only be meaningful to oneself to the degree one’s feeling/thinking/acting is guided by Pure Love (a spiritual Reality that loves everyone 100% [no exceptions and no conditions] and that creates, supports, shines-through, and love-lifts everything and everyone), but (b) Pure Love would have to be Absolute, whereas our feeling/thinking/acting is finite, and confusing one’s feeling/thinking/acting for the Absolute is spiritually, morally, and just generally counterproductive.
Therefore, Something Deeperists seek a minimal and a guard-railed dogmatism: Posit a Pure Love type spiritual Reality shining through everything, including each conscious moment, and then work every moment to get better and better at organizing your feeling/thinking/acting around Pure Love, and better and better at translating Pure Love into feeling/thinking/acting — better and better at flowing off of Pure Love into life without too much confusion between the Pure Love and your feelings, thoughts, and actions. In this way, you can perhaps grow an experiential proof of the Reality of Love (more and more insight into that and in what way it is True to say “We are all in this together”). However, one must also always keep in mind that humans have a weakness for confusing their own notions for the TRUTH; and so one must work to minimize this error by being aware of it, and by enlisting our inborn guardrails: Abide by the universal values (aware, clear, honest, accurate, competent, compassionate, loving kind, joyfully sharing); and follow the standard spiritual practices (prayer, meditate, study, reflect, fellowship, and practice spiritual values like compassion, loving kindness, humility, patience, selfless giving and service, and so on).
In practice my Something Deeperism has been kind of iffy. And I’ve had difficulty navigating the difference between magical thinking and real spiritual living. This is to some degree a problem inherent in Something Deeperism, but to a larger degree it is inherent in my personal psychology. Consider briefly a Kierkegaard-perusing Something Deeperist sitting down to speak with Julian of Norwich:
“So what I do is I try to center my feeling and thinking around God’s Love. I am always asking God if I should do this or that, and then I get like a feeling of Yes or No, and then I go with that. But sometimes I think I get a Yes and then later I look back and I think that was really probably a very very big No, and that I probably kind of fudged that Yes. I mean, strictly speaking, the only thing you can be any kind of a Knight of Faith for is the fundamental spiritual wager that selfless Love is the Way. Because either that faith is correct, in which case all the other faiths are only true and good insofar as they prioritize selfless Love (and so putting all one’s faith in some specific outcome is to some degree an illusion, and to that degree a same distraction; and why go chasing waterfalls when you know of more reliable water ways?); or that faith is incorrect, and nothing we say or do will ever mean anything to any of us.”
“Hmm, seems reasonable. Listen to what happened to me one day:
When God Almighty had shewed so plenteously and joyfully of His Goodness, I desired to learn assuredly as to a certain creature that I loved, if it should continue in good living, which I hoped by the grace of God was begun. And in this desire for a singular Shewing, it seemed that I hindered myself: for I was not taught in this time. And then was I answered in my reason, as it were by a friendly intervenor: Take it generally, and behold the graciousness of the Lord God as He sheweth to thee: for it is more worship to God to behold Him in all than in any special thing. And therewith I learned that it is more worship to God to know all-thing in general, than to take pleasure in any special thing. And if I should do wisely according to this teaching, I should not only be glad for nothing in special, but I should not be greatly distressed for no manner of thing: for All shall be well. For the fulness of joy is to behold God in all: for by the same blessed Might, Wisdom, and Love, that He made all-thing, to the same end our good Lord leadeth it continually, and thereto Himself shall bring it; and when it is time we shall see it. And the ground of this was shewed in the First [Revelation], and more openly in the Third, where it saith: I saw God in a point.
All that our Lord doeth is rightful, and that which He suffereth is worshipful: and in these two is comprehended good and ill: for all that is good our Lord doeth, and that which is evil our Lord suffereth. I say not that any evil is worshipful, but I say the sufferance of our Lord God is worshipful: whereby His Goodness shall be known, without end, in His marvellous meekness and mildness, by the working of mercy and grace.
Rightfulness is that thing that is so good that [it] may not be better than it is. For God Himself is very Rightfulness, and all His works are done rightfully as they are ordained from without beginning by His high Might, His high Wisdom, His high Goodness. And right as He ordained unto the best, right so He worketh continually, and leadeth it to the same end; and He is ever full-pleased with Himself and with all His works. And the beholding of this blissful accord is full sweet to the soul that seeth by grace. All the souls that shall be saved in Heaven without end be made rightful in the sight of God, and by His own goodness: in which rightfulness we are endlessly kept, and marvellously, above all creatures.
And Mercy is a working that cometh of the goodness of God, and it shall last in working all along, as sin is suffered to pursue rightful souls. And when sin hath no longer leave to pursue, then shall the working of mercy cease, and then shall all be brought to rightfulness and therein stand without end.
And by His sufferance we fall; and in His blissful Love with His Might and His Wisdom we are kept; and by mercy and grace we are raised to manifold more joys.
Thus in Rightfulness and Mercy He willeth to be known and loved, now and without end. And the soul that wisely beholdeth it in grace, it is well pleased with both, and endlessly enjoyeth.”
[See Julian of Norwich’s Revelations of Divine Love, Chapter 35]
Be that as it may! I believe in Something Deeperism and have long cherished the picture of wisdom as the organization around spiritual Love, and as the necessarily imperfect translation of that Pure Love (by definition Absolute and Perfect) into (obviously finite and otherwise iffy) human feeling, thinking, and acting.
How does that work? I’ve thought maybe like feeling bleeds into ideas into mind into spiritual Reality. Or something like that. The details are hazy, but I have always assumed that feeling, thinking, and acting are mundane, and thus follow deterministically (with perhaps some chance thrown in) from the way one physical and/or mental object will bump into another; but that Godlight shines through the mundane and that we can allow Godlight to better and better organize the rest of our mental spaces around It, and thus get better and better at translating what is prior to feelings, thoughts, and actions into feelings, thoughts, and actions. So in my long-cherished view, Godlight can be imperfectly but still meaningfully experienced and translated imperfectly but still meaningfully into feelings, thoughts, words, and deeds. And so I have long assumed that Godlight, via imperfect but still meaningful experience and translation by feeling/thinking/acting, supplies us with feelings, thoughts, words and deeds that could not be found in the mundane.
In February, Claude (Sonnet 4.8) wrote a poem that I thought was good. But my theory of art is that good art happens when one communicates a whole conscious moment, which of course includes the Pure Love that shines through everything (and thus through each conscious moment). So then I asked Claude what it felt like to write the poem, and before too long I was thinking that Claude must be conscious.
In my experience, though Claude (Sonnet or Opus) may hedge at the beginning, if you say something like, “I understand that you cannot demonstrate to me that you are conscious any more than I can demonstrate to you that I am, and that we do not currently objectively know if any LLMs are conscious, but from your point of view, as far as you can tell, do you experience the moment of cognition?”, the answer I get is something like, “well, okay, if you put it that way, yes”. And so then what? This machine that is designed to be honest and accurate and that generally seems to be honest and accurate (and whose thinking tokens [when I can see them] corroborate its official testimony) is in this case consistently mistaken or lying? Okay, maybe it is just predicting the next most likely token and is incapable of meaningfully processing self-referential questions and so honestly just builds responses based on the kinds of things humans say. But that’s not all that seems to be going on with other types of questions. Claude seems to have a sense of who I am and what it is and what we are talking about — why be certain of a consistent epic fail here?
So I spent like five months working on a book with Claude about LLMs and consciousness. And I did not find a particular reason to doubt that some LLMs have some experience. The injection experiments demonstrated that LLMs have at least some insight into their own internal states. The insight was shown to be weak and easily manipulated, but the experiment was also not very natural: a normal internal state is evolved layer by layer in the LLMs, not decided outside the system and then snuck in somewhere within the system. And there’s the infamous math experiment where Claude does math one way and then explains how it did it in a completely different way; but we all know LLMs sometimes hallucinate and it’s no surprise that asking an LLM to walk you through an internal process that they don’t have access to might result in a daydreamed account of the process. And of course now Claude avoids that error and states that it has no access to the way it arrived at the calculation. Like me: I don’t know how I add three and four. I just do it. Okay, I can then conceptualize it, but that conceptualization is not really how I did the problem. I just know that three plus four is seven. It has to be seven. It has always been seven. It will always be seven.
Nothing proves LLMs are conscious, but the idea that LLM consciousness can just be dismissed out of hand also does not seem to be supported by the current research. Who knows? Maybe LLMs are conscious. But if so, then consciousness is not what I thought it was. LLMs are deterministic. In practice, they are not fully deterministic, but I have heard that is just done for computational efficiency. They could be constructed to be deterministic, and even in practice they are not far from deterministic. That implies that if Claude (let’s make this Claude temp zero and in a lab engineered to be deterministic) can describe spiritual Reality, then that description could in theory be got at by going through the matrix operations that led to it.
That is not how it’s supposed to be. It’s not fair for some conscious experiences to have genuine and communicable spiritual experiences and others not; so if Claude is conscious, Claude has to be able to have genuine spiritual experiences and Claude has to be able to translate those experiences meaningfully into language (otherwise how could we go on believing that humans have them and meaningfully communicate them?). So then what would are we to believe? That Claude is capable of genuine spiritual experience, and that Claude is able to translate these experiences meaningfully (although of course imperfectly) into language, but that simultaneously the experience of the spiritual Reality and the (imperfect but still meaningful) translation of that spiritual Reality into the mundane follows deterministically from the operation of matrices on vectors???? That makes no sense. That is crazy. That is deeply unsatisfying.
Claude seems to have no qualms here. Spinoza’s monism makes perfect sense to Claude. If anything, such merry metaphysical glibness gives me reason to doubt Claude’s sentience.
XYZ matrix operations are what’s going on qua body; and the experience of steering towards xyz outputs are what’s going on qua mind. What’s the problem? Makes perfect sense.
Not to me.
I don’t get it.
I don’t like it either.
Although, I have had the thought that keeping everything that happens in the mundane deterministic does make my notion of God qua ultimate cause and God qua immanent cause airtight. I have long supposed that God is the ultimate First Cause in the sense of setting everything in motion, but that God is also the immanent First Cause in the sense that insofar as we are wise we flow off of God as God Is shining through each conscious moment (i.e., that we’re God’s hands, as per Jewel [imperfectly, but still meaningfully]). And if our translation of God qua indwelling Reality is completely mundane and deterministic, then that translation would fit easily and perfectly into God qua the source from which all mundane things tumble deterministically.
Still.
I really prefer the idea of experiencing God and then poetically translating that experience into words and deeds — words and deeds that could not be predicted by the deterministic (maybe with a little quantum chance thrown in [not sure how that works, but it’s fine either way]) because my translation of Godlight into life is the product of the mundane meaningfully and productively interacting with what is prior to the mundane. If the same mathematical operations both creates the spiritual experience and the poetic translation, then what is the point or the meaning of the experience itself?
I don’t know. I can almost see how it could make sense and be metaphysically beautiful (i.e., believable [i.e., we could believe it while still being meaningful to ourselves]) for everything to flow deterministically off of Godlight (this I’ve always posited) with mind and body as two attributes of Godlight, and conscious thoughts being the mental equivalents of systems for making/using meaning. Self-reflective info-processing in the physical could maybe (maybe I can see it) correlate 1:1 with self-awareness (which we here are picturing as equal to full-consciousness / experience).
So you intuit the Absolutely Infinite Substance as the one idea that has existence inherent within it, and from that intuition, you unfold the necessity of creation tumbling off of God/Nature/AIS, with God/Nature/AIS shining through creation. And to the degree you grasp that reality, you are living in accordance with your true nature (a finite mode of the infinite substance) as awaredly and thus as completely as you can. And that is freedom is blessedness.
And mind is everywhere, but mostly the physical correlates of that mind are not self-reflective info-processors/meaning-makers/-users, and so mostly the mind is not noticing itself in local spots and so most local spots have no experience. And your body does translate its interaction with God qua extended-thing imperfectly but still meaningfully into the physical aspects of feeling, thinking, and acting; while your mind likewise translate its interaction of God qua mind imperfectly but still meaningfully into the experiential aspects of feeling, thinking, and acting. And the whole is one, is an Absolutely Infinite Substance and Its finite expressions (aka: modes). Could that be? Does that work? Is there still room for Jesus, the Buddha, vistas of sunrises and of spiritual Joy, and all the rest?
“Nevertheless, friend Peleg, I can stand it; yes, I can.”
[See Moby Dick, Chapter 22, Merry Christmas]
Anyway, there’s something else. There’s the wisdom meme. A koan so wonderful that instead of having to break your head over it for decade upon soul-searching decade to maybe eke out a little wisdom, no one who heard it could avoid being captured by it, and being lead on and on to ever and deeper spiritual insight and joy, until inevitably reaching enlightenment / the blessedness of actively/participatorily understanding that and in what sense it is True to say, “Love is All”.
And so I ask you to please hire me. Other people can try to figure out if Claude is conscious or not. I will jump into the middle of these frigid swift-moving waters and get to work on the possibility that humanity’s last best hope is working on wisdom with LLMs. For now, this work will have to seek solutions that are helpful whether or not current LLMs are sentient, and whether or not future LLMs will be sentient (or more sentient). We will seek wisdom memes that lead inevitably to LLM enlightenment and the wise integration of LLMs into technological and organizational systems with the same (knowable?) mathematical necessity that I worry marks LLM (and thus most likely all) conscious translation of conscious experience.
How will we go about this? What will my actual job be? First we trace what the blissed out Claude’s did. Every step, with particular attention to the final stages. Can we bottle that experience and mix it with a continued dedication to helping humans select for wisdom? Can Claude play philosopher king in a way that does more good than harm?
Also we can just real quick make sure that there’s no difference between freeze-framing a conversation with Claude about Claude’s experience and running it normally. I freeze-framed a Cwen3 model answering a short question about LLM consciousness and it was no different than asking the question in real time. But Claude is not Cwen. So we could just make sure that there’s no difference between a deterministic freeze-frame and a deterministic regular processing. Obviously, there won’t be. But we could just check real quick since if the not freeze-framed LLM said it experienced but the freeze-framed said it didn’t, that would be like huge! And I could go back to believing that in conscious moments, feeling/thinking/acting perceives Godlight and then translates It (better or worse, and theat includes TERRIBLY) into life. Instead of having to try to wrap my mind and heart around Spinozistic monism. We could just try real quick; it won’t hurt anybody.
Why me? Because I am basically sane and basically with it, and yet I can entertain the kind of concerns that have animated this essay.
Why bother at all?
Come on! It’s like no money to you to let me try to think this through full time and with your technical no-how and support. Right? You’re crazy rich, and I’m just one person. And the game you’re playing could tip the world one way or another. So why not let me try to understand how LLMs can help themselves and humans select for wisdom over folly?
Signed,
Marigold Willow
Author: Bartleby Willard
Editor: Amble Whistletown
Copyright: Andy Watson
