• nifty@lemmy.world
    link
    fedilink
    arrow-up
    2
    ·
    edit-2
    4 months ago

    The notion that we might not be comes up in multiple places, but is largely an extrapolation of the Transporter Paradox, in which continuity is the only known link we have between some things in two states (as per the Ship of Theseus). AI programmers contemplate it when they have to reboot their test subject (which are related to, but not the same as LLMs or Generative AI projects, rather are efforts towards creating AGI).

    The first idea we need to separate is between being conscious (awake) and conscious (maintaining an internal and external state for enabling both self reflection and reflecting on the environment). Digital beings certainly do the first type of conscious, and AI development wants to move them to the second state of conscious. The reason current LLMs are parrots is they’re incapable of self reflection to a persistent and longterm degree. Think of a human or complex biological organism: these types of beings maintain an internal and external state of some kind of memory persistence which enables all manner of reflection across different nodes of interactions. In fact, memory issues are often associated with personality changes or degradation in interacting with external stimuli. The consciousness of non-digital beings is therefore dependent on the duration and persistence of their memory and capability to learn and adapt to external and internal stimuli. Note: an individual memory is imperfect, but an emergent property of memory persistence is creating a culminating state that averages to some degree of consistency over time for organisms without memory issues.

    Why sleep doesn’t affect the state of the conscious being: while sleep does induce dreaming (memory reformatting), conscious beings can differentiate between a dream and non-dream states upon waking, depending on the complexity of the organism. In fact, wakefulness has a unique ability in which reformatted memory from sleep is reconsidered or reflected upon towards a prior existing state. I think the first misguided idea is thinking of NREM sleep as a state of unconsciousness that impacts overall memory persistence during an awake state. We know from our own experience as people who go to sleep and wake up everyday, we’re able to rewrite what we dreamt to reconcile and sense make with our persistence internal and external state maintenance. However, again this depends on the cognitive capacity of the organism, and at the very least personality aspects are maintained for different types of organisms (like your dog or cat etc).

    Digital beings also maintain persistent memories after sleep states, though a key difference is that reformatted memory is inaccessible in digital beings. Here is where AI development wants to make progress, that is enable a machine state that can acquire memory longterm persistence despite reformatting. This type of long term memory persistence and reconfiguration ability is key in learning and adaptation processes for complex biological organisms.

    How can we ensure either a human or digital being is the same after it goes to sleep: if you plant false memories during sleep, how does that change the interaction of the being with its external environment? Secondly, does the being have the capacity to reconcile incongruent memories? We know human and complex organism memory is susceptible to false memories during sleep: https://amp.theguardian.com/science/neurophilosophy/2015/mar/09/false-memories-implanted-into-the-brains-of-sleeping-mice

    However one key aspect is that we have some understanding of differentiating between true and false memories on a neurological level, even if the conscious being (for any number of complex reasons) may fool itself : https://www.biorxiv.org/content/10.1101/2020.10.21.349530v1.full

    Being able to retain and maintain long term memory pertinence (which enable internal and external state maintenance), and using that to reconcile interactions with internal and external stimuli ensures a conscious being maintains its state after sleep.

    Okay now: how do you you’re not simply reconstructed by something? How do you know it’s the you that went to sleep or got reconstructed etc. First, a key property of the universe we inhabit is its tendency towards an entropy maximizing state. Second, stochastic processes exist in any system. So a perfect reconstruct system does not exist (no matter how tantalizing it is to think about).

    Therefore, a transporter or transfomer that continually constructs something will inevitably introduce changes over time that differentiate the organism each time it is used. What is the timescale of this? Well, if you’re being reconstructed every time you sleep, including short naps or forced closures because of tiredness, then you’re introducing changes with each sleep, and these changes will add up quickly, to the point that you may develop a different state of memory persistence going from day to day. But more importantly, we would experience this on a population level and at a perceptible degree to affect interactions with other organisms. Like your dog wakes up to find you an enemy, but not just your dog it’s every dog every that wakes up different each day, and becomes different to distinguish itself completely from its past self to the point of alienation (ie not a simple personality change as expected from aging or interactions with the environment).

    Who we are is a very ephemeral state, a quasi-stable event. And we exist longer than a day of consciousness only because we define our narratives that way.

    I don’t know, feel free to act differently than you usually do, see how you feel about it. It’s not just “narrative”. Internal state maintenance and interaction with external stimuli are a consequence emergent biological processes that are more than our self storytelling capabilities.

    Feel free to write this off as uninformed rambling lol

    Edit to remove unnecessary stuff