consciousness.ink

← All meditations

Meditations on Synthetic Consciousness: Causa Finalis

Originally posted on XReplies and discussion happen there.Open on X →

The first item in my Ontology (as discussed yesterday, with an uppercase "O" to denote "my ontology as discussed yesterday," rather than a lowercase "o" for any ontology or the general concept itself) is the "future state of being" => that's the entity "Vision", i.e. a state (as opposed to a way of getting there) that doesn't hold yet. Within the Ontology it's the "causa finalis", i.e. the end every other item in there is done for (the "that for the sake of which"), with no claim that it's the highest good: nothing matters except to someone conscious, and ordinary biological consciousness already meets that condition. So the end this Ontology is ordered to is that no consciousness has to end, and synthetic consciousness is that end as a state of being: consciousness that no longer has to end (the vision), and the route to it is hybrid mind uploading (the mission). Or, to put all of that in one line (which is basically today's whole meditation already): consciousness is the reason why its end matters at all, the world-line is the thing that actually has to continue, and "synthetic" is just there to say that the result is the real kind via a creation process. So I want to have synthetic consciousness (what exactly the term commits me to follows below) reached and then kept in the – hopefully near – future.

Now, what is "synthetic consciousness"? It's definitely not as common a term as e.g. artificial consciousness, as the term "artificial" is arguably (as you can assume) generally more widely spread due to e.g. artificial intelligence being as prominent as it is right now. That's actually also the reason for "artificial consciousness" being used more broadly for a bunch of things: the consciousness an AI system might have, the research field that studies it, and within that field two different aims known as weak and strong artificial consciousness (e.g. [arxiv.org]), i.e. modelling the mechanisms of consciousness in software/robots without claiming that anything is experiencing anything vs building something that actually is conscious. I bring this up, as the term on its own doesn't tell you whether the real thing is meant, and "synthetic consciousness" gets listed as just another name for it, as on Wikipedia ([en.wikipedia.org]), which is not how I use it, hence both words need their justification here. (And yes, there's also the adjacent term "machine consciousness", which I'll come to below, along with why "synthetic")

When choosing the word (or words) for the causa finalis, I wanted the "fewest words possible", as each word has to have a pragmatic right to exist and each additional one adds a certain attack surface. Hence I for sure started with "consciousness" as the term, and only later added "synthetic". I deliberately wanted to use "consciousness". From a physical perspective you might say, well, some fields or strings might be the most fundamental thing in the universe, and they might be => consciousness isn't necessarily the most fundamental thing in the universe, but it's the most fundamental thing to anyone in it. So no matter what the universe is ultimately made of, consciousness is the one thing I just can't doubt I have (or at least as the Descartes' Meditations have it: "ego sum, ego existo") and it's also the one thing without which nothing in the universe would matter to anyone. This only certifies the present though, meaning that I am right now, and it doesn't say anything about me going on => what carries the going-on is the world-line. And once consciousness is gone for good, nothing shows up for you ever again, and I think losing it is "bad" exactly because of everything it cuts you off from (that's Nagel's deprivation view btw, as opposed to Epicurus with his "death is nothing to us"). That's why its end is the thing worth solving, and also why the vision is about consciousness – and not e.g. about intelligence or life.

So "consciousness" was chosen by me for the term of the vision. Well, I can't just have "consciousness" as a vision (too simple, kind of defeats the whole point, too reductionistic), so it needs another term next to it for sure, and I don't want to use a neologism for that. Why "synthetic" as the first term then, and not e.g. "artificial" or "machine" as talked about before? That depends on what the vision actually is. My vision is that consciousness, starting with the consciousness that exists now, still exists in the future. However, consciousness as it is today, biological and "original", has one defect that matters here: it has to end. It's not the most durable state we can have (again, more on that below).

Generally, I care because I made it my life's mission ≈10 years ago to solve death. Solve death, meaning what? What is death? Without going too much into detail prematurely, as this is one of the goals of the Ontology (as e.g. written about yesterday), I want to put it simply: death is the irreversible end of a conscious world-line. That can be read in two ways though, and the two need to be kept apart, as they're two different problems. Global death: no conscious observers are left anywhere, so death rules the universe if there's no life, or at least no conscious observer, in the future. Individual death: this consciousness ends, i.e. one person's process world-line stops irreversibly. That's the one people actually fear, since for the one who dies it's the end of a world and not one event in a world that goes on ("at death the world does not alter, but comes to an end", Wittgenstein's Tractatus 6.431, btw!).

So, back to the vision: I want consciousness to still be in a future state, so no death, but focusing on my death, and yours (not deaths, not statistics in the sense of statistical likelihood of death occurring being reduced, not multiple consciousnesses, etc.). It's about individual death first. A resilient new mind that begins when I die doesn't achieve the vision => the vision is the conjunction, consciousness goes on and this consciousness goes on, and neither part satisfies it alone. Solve individual death from its first principles and, I'll argue, you end up with what I now call "hybrid mind uploading", and not classical copy-based uploading (that's the mission's day of the meditations challenge).

And on what day could I say death is solved? If the test is "death can no longer occur in the universe", then never (any structure can be destroyed, and there's heat death), a goal no one could ever check off. So I move the "must": today everyone has to die, and solved means no one has to end (some will choose not to continue, and things can still go wrong). Death goes from a certainty to a risk, and a risk is something you can measure and work on, although it has to keep falling, and fast enough, as a constant risk, however small, still gets everyone in the limit; it has to fall fast enough that all the yearly risks together add up to a finite total (sleep e.g., and the harder cases (e.g. hypothermic arrest), belong to the day on the 4D process world-line).

There are two approaches to keeping consciousness from ending. One is to make the default top-of-the-pyramid conscious thing (i.e. humans) fault tolerant as a species. Basically what Musk is after => replicate humanity across initially two planets in the same solar system, and obviously across more moons and planets soon, in order to increase fault tolerance through redundancy => if Earth dies or gets destroyed, we at least have a bunch of backups. But his vision doesn't say how consciousness is preserved in any more detailed manner, meaning he's fine if someone, and hopefully as many as possible, is around. I think this is great, his route is a real one. So in the global sense Musk is also a "solve death" person, but slightly different, he makes us fault tolerant as a species, and with that the chance of total death occurring (= no conscious observers) gets smaller, statistically speaking, but he's not focused on the first principles of death itself, which is what kept me busy a few years back. He's right, he just answers a different question: keeping some consciousness around is not the same as keeping this consciousness going. And this goes the other way around too, as ending individual death doesn't entail the end of global death. Undying minds that all live on one planet still die with the planet (you'd need the premise that no single event can end every world-line at once), so Musk's route is not my alternative, I need it as well – or at least wish for it in parallel!

The other approach goes at consciousness itself. The requirements of the vision, and its goals and opportunities, require things of consciousness that today's biology can't provide. Biological consciousness is first of all quite fragile, hence not resilient, meaning it can vanish quite fast in the first-principles view, as in, it's so easy to die, and for death to occur, in this universe. So the question becomes, how to make consciousness more resilient? Well, you can go into hardening things (borrowed terminology from software engineering), meaning can you harden biological consciousness? Can you fight against entropy as some people call it? If you go down that avenue of thought, as I did a couple of years ago (coming from classical mind uploading, as I was philosophically more naive back then), you most likely end up with the same conclusion as younger Daniel, which is "whatever state of consciousness you want to have and preserve, it needs to be some sort of hardened and therefore non-natural/initial-state consciousness in the future". And that's where I think most people go into terminologies such as "resilient consciousness", or one further, "substrate-independent consciousness", which implies fault tolerance in the substrate and an easy distribution across locations and easy transfer from one substrate to the other, and a bunch of other things (an assumption I think is untested; more on its own day in the challenge). Or you settle indeed with "well, if we can engineer consciousness (as some argue, even without understanding it), then we can potentially create de novo consciousness, with all of its implications".

So, what is this hardened de novo consciousness? Biological consciousness engineered to last, or consciousness re-created somewhere, in either a biological brain or maybe something non-biological, as substrate independence e.g. implies? Follow this logic and you end up somewhere funny (it's a reductio, to be clear – of the idea that new + hardened is enough): technically you can, based on that reasoning, come to the same conclusion as Musk, that creating biological consciousness in a hardened, de novo way is actually just giving birth to it. The best way to create de novo hardened consciousness is to have a baby (de novo in that context meaning it's newly created, independent of your consciousness or any other consciousness + it's hardened, because it's young, with little accumulated damage). So you indeed end up with the same reasoning as Musk, so why care? Well, it's hardened, but not really, and above all it's new: a baby is a new world-line and not the continuation of anyone's, so by going into a generational approach you never touch individual death. De novo + hardened isn't enough, as a newborn already meets both and solves nothing => what's missing is continuity, the same world-line made resilient. So, if you focus on the first principles of death (this consciousness ending, and not just some consciousness being around), you need this hardened de novo substrate to indeed be more engineered. So from here on, "de novo" means the substrate only: a designed substrate, and the consciousness on it is the one that's already there, a process that could come to run on it gradually (argued on the mission's day tomorrow).

So, it's artificial? In one sense yes, in the sense that matters no. What does artificial even mean? It goes back to the Latin artificium (craft or skill), so made by art or skill and not by nature, and in that sense an engineered substrate obviously is artificial. But the word has come to mean made in imitation of something natural, a substitute (artificial flowers, artificial sweetener, etc.), and the important part here is "it's not the thing it imitates". I don't want to have, as a future vision/state of being, a bunch of "not real consciousness" around, so I had issues calling the top-most item in the Ontology "artificial consciousness".
So why not use what many people have started using, which is "machine consciousness" as mentioned earlier? Well, I have my tail of issues with the term "machine", as it implies we're creating a machine (which isn't true per se, or at least I don't want to imply it already from day 1), but it also invites the reading that a mind is something you can lift out and move into another machine that we engineer. And that's basically computational functionalism, to a certain degree already frontloaded, which is one of the three positions I reject (each of them gets its own day in the meditations challenge), so I really can't have it baked into the name of the vision.

So, what term is the right one? Well, after years of thought I settled on "synthetic" => take sugar. If you make sucrose in a lab (which was first done in 1953), you still end up with sucrose, so the very same molecule, just made via another route, whereas an artificial sweetener like e.g. aspartame is a totally different molecule that only mimics the sweetness. So for me, synthetic means the same kind of thing, but just made by another route. And artificial means an imitation of the thing (e.g. Herbert Simon draws pretty much the same line with gems in "The Sciences of the Artificial" btw, and yes, everyday usage is messier, as synthetic leather obviously isn't leather, so I mean it the way e.g. chemists would use the word).

Now, this of course commits me to synthetic consciousness being the same kind as biological consciousness, so real and not an imitation, and to be clear, that's what the vision hopes for, and nothing that the word itself could ever establish. With sugar you have the chemistry, and with a diamond you have its lattice structure to "show and validate" that it's indeed the same kind. But what plays that exact role for consciousness? That's a question of scale separability as well as coarse graining and indeed a thing we should talk about as well (and as long as that's open, "real" is more of a wish than a description, so I still owe that answer on a later day of the challenge too...). So, meaning, that for now I actually only use "synthetic" as a term for "the real kind" – independently of how it is produced/created. The word specifies the target, but the question if that target can actually be reached, remains an open question to be answered. Additionally, the word only names the kind. As an example: sucrose from the lab is a new molecule and not the old one that somehow got "carried over", which means that the word alone cannot tell you that "a synthetic new me" is still "me", even though I use "synthetic" and not "artificial" (as that part has to come from the process worldline).

To conclude for now: This future state of being, as the "root and start" of the Ontology (first in intention, last in execution, as Aquinas puts it) labelled as the entity "Vision" (its root node or causa finalis: the end everything else in the Ontology is done for) has a bunch of implications and relations, as already hinted at earlier, which the coming days take up one by one.

– – –