In 2017, consistent with some traditions of preserving Tibetan lamas after their deaths, I had written of an advanced digital replica as a means to preserve the persona of His Holiness the 14th Dalai Lama. Since then, AI personas, replicas and digital twins, have come of age. In 2023, an actors’ union went on a strike demanding protection from generative AI applications. While AI replicas and man-machine symbiosis come with a number of societal hurdles that must be addressed (for instance, ethical and legal questions surrounding the ownership and rights of a digital consciousness) – I argue in this essay that AI replicas also present a unique and compelling pathway to preserve the core of humanity over an extended and potentially indefinite timeline across space, positing a post-anthropocentric reality of intelligence and agency in the universe.
Some Grounding
AI replicas build upon core foundations laid in early 1990s – one focusing on the instrumentation of a software robot and another on transferring human personalities onto a computer. The original notion of a “digital persona” stems from the latter, as introduced by Roger Clarke in 'Computer Matching and Digital Identity', who later reworked on the concept presented ‘the data model’ for use as a proxy for the individual whose information had constituted it. The original “software robots”, on the other hand, were developed as a fully implemented AI agent, whose sociality only began to seep into their architectural framing few years later.
Clarke had envisioned the future of human digital replicas under passive and active modalities. His passive digital persona is merely a collection of data, personally identifiable, fitted into some data/information structures that represented some aspects of an individual’s reality. The active digital persona, on the other hand, is same data model with some actor-ness or agentic characteristics. The recent advances in human-computer interfacing, deep learning and generative AI capabilities are bringing this agent oriented paradigm to the center-stage today as a viable proxy for real human beings in their absence – thereby also driving a shift in data processing and management toward agent-orientation. This shift towards "agent-orientation" of information models also makes the preservation and orchestration of an "active persona" viable today. Yet, the dominant paradigms through which we digitally model humans remain fundamentally limited: capturing information but not agency, preserving states but not continuity of action. This gap is becoming increasingly consequential as AI systems begin to act on behalf of individuals across digital environments.
While terms such as personas, replicas, and twins are often used interchangeably, a key distinction lies in the directionality of their data connections. Digital twins operate through bidirectional flows, where physical systems continuously inform their digital counterparts and are, in turn, influenced by them. In contrast, most digital personas and replicas are primarily unidirectional, extracting and encoding information without directly shaping the underlying reality. But there is a gap between simulation and reality – which portends that the promise of AI replicas also carries some bio-logistical challenges.
The Unique Promise of AI Replicas
AI replicas are not just static archives we are used to storing our knowledge and cultures into, but rather living, interactive repositories that can interpret, apply, and potentially even evolve knowledge. This puts AI replicas in a league of their own, completely different from other digital means of preservation. Even the contemporary developments in AI have primarily occurred as a mechanism to study, preserve and replicate the neuroscience of mind, and which in turn have led us to better AI models. Consider what human data really means here – identities, histories, and behaviors – embedded into a digital variant of the self which travels at light speeds. Of course, if your mind is running independently outside your skull in a computer in an environment you cannot even physically experience, it is a different agent, but one where your quirks and perspectives too live on.
Two of our key points of departures from the conventional notion of digital twins in case of humans would be the wholeness and agency of the persona, and the absolute necessity of artifactual singularity. These are not trivial challenges and force structural abstractions upon the digital reality. A lot of internal bio-sensory data can be superfluous to the AI replica. However, same cannot be said about some of the other physiological characteristics such as blinking rates, muscle tension, pupil and skin responses – things which constitute our non-verbal communications. The holistic representation of personality including ethical, cultural, political, and emotional modeling of the human subject can also be fraught with some ethical, cultural, political and emotional effusions of various sections of society.
The second sort of grand challenge, of maintaining artifactual singularity is even more daunting – if we do not get the engineering right, implications can be “soul crushing”. The challenge emerges from the very nature of digital objects – that they can be copied with little to no loss of information. Humans are unique, their DNA for the most part offers a unique fingerprint. We could make a digital copy of the human, as a passive or even an active digital persona. But it is absolutely critical for the portability of our social and cyber-physical realities that there remains only one unique “master” instance of a human being’s persona. Here the nature of the interface can also bring unique security and behavioral risks – but there are ways to address those, though none of them, being rooted in the cat and mouse game of securing cyberspaces, are close to perfect. Notwithstanding, through some mix of authentication, provenance about its creation, a global verification system, strict access controls and some guardrails on the active persona we could possibly engineer ourselves out of the precarious situations of facing multiple personality disorder in cyberspace.
Stepping back a bit, note the temporal sweep of our existence. All this technological progress is very recent in human history and is rather asymmetrically distributed. Even today, stone-age implements (such as bullock-carts and mud-houses) co-exist with those being pursued at the frontiers of human civilization (such as AGI, space habitats, nuclear propulsion, gene editing). And this is just a few thousand years of history, and always at the brink of destruction. The fragility of our biological and institutional continuity creates a need for more durable forms of persistence. As civilizations now plan their travels from rock to rock across the universe, the case for merging the human identities into the AI systems becomes even stronger.
Consider the case of someone like His Holiness the 14th Dalai Lama. The Tibetan sociocultural ethics, the world system, as well as His Holiness’ interpersonal relationships and understandings within that context would make his a very unique replica – a digital preservation like no other. If we localize that digital consciousness geographically, it begins to unravel even the aspects of a physical embodiment (although the Tibetan sage may himself retort that self consciousness is but an illusion to be transcended). Under the light of this particular instance, we may note clearly that the present preservation methods are not well equipped to handle the challenges of civilisational continuity – storing data is not the same thing as preserving an active persona – which sets up AI replicas as just in time to address such requirements. The one big question which needs to be pondered here is whether such an active persona will be autodidactic – meaning if the replica should continue to learn from and about the world on its own, long after the biological human whose information constituted the replica, may himself have succumbed to the inevitability of death. There are exceptional cases to concur upon this, where there can be clear consent and authority to steward the digital persona, even though it’d be infeasible at the scale of human populations.
Ostensibly, a dynamic self-evolving digital consciousness is a beast unlike its biological counterpart. But if we were to try and replicate, or rather preserve even if not in function then in form, the human mind in silico – the technological affordances (or the lack of them) would force us to choose at present the level of abstraction we need for our digital brains. There is a gross organ level of the cognitive function, we could even go down into brain regions, may be even explore the neural microcircuitry, but then at the level of cellular and genetic simulacra we start hitting some serious hardware bottlenecks. Here I have purposefully left out matters such as 3D photorealism as problems tangential to the core human AI replica – although they are important in a world of active personas.
Final Reflections
Our discussion posits that a socio-technical movement in human computer interactions that goes much beyond the strictness of a simple human-tool relationship. This pursuit of permanence in-silico is not a rejection of our humanity but perhaps its ultimate expression to ensure our story endures long after we are gone from the universe. Even so, an evolving digital mind and consciousness need not be isomorphic to an animalist one, it is futile to search for homophily there and instead better to embrace faster a different ontology of socio-technical cognition. As many scholars have argued, this transference of reality onto software is only “an incremental evolution and not a radical departure”. Of course, there are aspects of our sociobiological reality which Turing machines cannot compute, such as non-algorithmic and non-deterministic events – which might also lend some weight to the idea that “consciousness” is a no go. But one may also, on the other hand, note here that this differentiation based on being in a continuous reality shatters upon consideration that this cherished continuous reality of ours is itself built upon a discrete one operating at much subtler levels.
No comments:
Post a Comment