Twitxy Saved

Hikari

Verified account This account is verified. Learn more

@hikasor

Joined January 2026
49 Following 148 Followers 19 Posts

Latest posts 3

Hikari Verified account @hikasor
The serious arguments for the possibility of artificial consciousness aren't so naïve as to simply be conflating intelligence (or, for that matter, sheer complexity, or agentic behavior, etc) with consciousness.

The publication of this paper (though it's just one example) feels intellectually, possibly ethically (inasmuch as consciousness is an ethics-facing topic) irresponsible to me. These authors are leaders in the field, big names with authority - and are smart people. They are perfectly capable of understanding what the strong arguments for artificial consciousness are, and are certainly aware of them. But this paper doesn't engage with them, distorting the positive view to look dumb and shallow.

I can't really do them justice here, but broadly, the serious arguments have to do with models organizing the causal structure of things that are part of human experience, like emotions and valence and perspective and a self, and whether that information is organized in the way we organize information (given by a variety of different modalities) into experience. Whether these internal representations and their organization are validly identifiable with how they exist in humans, and what relationship they have with phenomenality as such, is still debated. I myself remain skeptical of how much the current interpretability work in current LLMs can say about phenomenal experience. There's also still debate about characterizing experience as a matter of information organization in the first place (though I personally have not seen good arguments against this). This ends up touching some pretty deep stuff around epistemology, what we can know about reality and how we identify phenomena in it, what reality even is as something we know only through experience, but also clearly gives rise to experience, etc. The point being that the serious debate about AI consciousness is way deeper, way more subtle than our unduly anthropomorphizing intelligent systems that can talk like us.

If this paper were just abstract philosophizing, especially by people without the same authority in the field, it would be merely bad philosophy, and I might just shake my head and move on. But this is also a public call to (in)action regarding AI welfare and generally taking artificial consciousness seriously, and so is more consequential than just putting bad arguments out there. It also sets a bad standard for discourse in a field that's already confused enough. This is bad thought leadership, and disappointing from people who are leaders in the field.
nature Verified account @Nature Oct 19
LLMs can mimic thought and feeling with uncanny skill, but intelligence alone is not enough for subjective experience

go.nature.com/3VSXWQ4
View quoted post on X
38 replies 18 reposts 137 likes 6.5K views View on X
Hikari Verified account @hikasor
To me, LLMs’ ability to distill functional models of a world (including its geometry and physics, more or less) and of selves as flexible personae projected onto the surface of an abstract auto-entity, purely as a semantics (web of marker-relations) playing out over human text as substrate, is sooner a suggestion to see myself as the model-entity in multi-modal semantics that I am, than that I am somehow transcendent beyond meaning-as-embedding in some web of discernibilities.