iiinova — where your digital self keeps evolving

Paper (Philosophy & Technology 39, 151): doi.org/10.1007/s133... Implementation paper: arxiv.org/abs/2608.02986 Code: github.com/ksk-S/intern...

doi.org

Client Challenge

000

The implementation has caveats of its own. The agent still lives on Devnet; its "environment" arrives via signed oracles; lose the key and it's gone. And its drive — economic metabolism — was given to it by me. Norms of its own, intrinsic normativity, are the next frontier.

100

Limits remain. Blockchains run on electricity and nodes; forks and regulators exist. "The absoluteness of code" is an ideal type. And whether this agent 'experiences' anything, the framework cannot say. Even so, the fact stands: an outside can be built.

100

Its survival rests on three implemented requirements: 1. Individuality — one key, one unforgeable identity 2. Continuity — every state chained; no forks, no tampering 3. Economic metabolism — pay fees or perish Continuity of a new kind: secured by cryptography, not metabolism.

100

At every step the agent proves, in zero knowledge, that the right individual, in the right body, computed the right thing — and a frozen on-chain verifier (one even I can no longer change) checks it. No proof, no transition, no matter who asks. No impostors; no rewritten past.

100

Internalising the Identity Primitive: Cryptographic Individuality for an Autonomous Agent on a Public Blockchain

So that this wouldn't remain speculation, I built one: a cryptographic individual living on Solana Devnet. Its core is a single key. A neural net derived from that key senses its environment, acts, and lives autonomously on-chain. arxiv.org/abs/2608.02986

arxiv.org

A software agent on a public blockchain accumulates authority and economic stakes, raising the engineering question of what makes it count as an individual. The paper's central contribution is a shift...

200

The other upshot concerns consciousness. What we count as real quietly sets what we count as sentient — and who gets moral status. Biological naturalism, computational functionalism, social constructionism: the deadlock between them mirrors a deeper clash over what reality is.

One upshot is for the ontology of agency. Artificial Externality makes possible agents with cryptographic individuality: individuals that persist and resist interference by key alone — and that point toward artificial agents with a kind of continuity the world hasn't seen.

Kant's thing-in-itself can't be produced. Latour's irreducibility can't be designed. Harman's objects withdraw from every relation on their own. What none of them anticipated: absoluteness that is engineered. We can now build the absolute.

100

The heart of the paper: Artificial Externality. Deploy a smart contract and it outlives its maker's death, its company's bankruptcy — so long as nodes run. Unstoppability is the design goal itself, enforced by cryptography and consensus. A human artifact, indifferent as nature.

100

The paper sorts reality by how things resist our mastery. 1. Substrate: matter outruns any complete description 2. Contingency: experience shrugs off the verdict "this is fake" 3. The Inexorable: autonomous systems defy intervention Behind all three: frustrated mastery.

100

What is real? In the 18th century Samuel Johnson kicked a stone to "refute" idealism: the pain in his foot, the stone's stubborn push-back — that was reality. But amid deepfakes, VR, and crypto assets, kicking stones no longer settles the question.

100

Artificial Externality: A Three-Layer Model of Reality from Substrate to Smart Contract - Philosophy & Technology

I've written a solo philosophy paper. link.springer.com/article/10.1... What is real? I argue we should understand reality through resistance — how things push back against our grasp. And smart contracts now give artifacts the kind of "outside" that once belonged only to nature. 🧵

link.springer.com

What is real has always been something we find, not something we make—or so philosophy has assumed. This paper argues otherwise. Characterizing reality through resistance rather than substance (the wa...

2122

Reposted by Keisuke Suzuki

Manuel Baltieri @manuelbaltieri.bsky.social · 10/07/2026

New work led by Filippo out on Neurocomputing! W/ @ksks.bsky.social , @kanair.bsky.social and me. Does an agent need to be "aware" of its executed actions to learn to achieve a certain goal? In reinforcement learning, the answer is generally "yes". 1/

151

Keisuke Suzuki @ksks.bsky.social · 01/07/2026

very first night in ASSC in Santiago

0110

Keisuke Suzuki @ksks.bsky.social · 30/06/2026

are you not here this time?

100

Reposted by Keisuke Suzuki

Models of Consciousness @moc7.bsky.social · 11/06/2026

Meet the Program Committee behind this year's Models of Consciousness conference: @joannaszczotka.bsky.social , @ksks.bsky.social , & Robert Prentner! Their expert curation bridges mathematics, neuroscience, & philosophy to advance rigorous consciousness science at hashtag #MoC7

195

Reposted by Keisuke Suzuki

Antonino Greco @agreco.bsky.social · 20/05/2026

cool paper on computational phenomenology with a unified framework on perceptual hallucinations! 🧠🍄🦄 using deep neural networks to generate testable predictions about altered perceptual phenomenology #NeuroAI 👇👇👇

041

Thanks to editor Axel Cleeremans and reviewers Renato Paredes & Jessica Harding. Comments and critiques very welcome 🙏

000

10/ Bottom line: phenomenology shouldn't reduce to a single principle. Different objective functions × architectures generate qualitatively different experiences. CP becomes an exploration of design space — not a reduction to one master equation.

130

9/ Extension to ego dissolution: lowering D's threshold in the multisensory domain corresponds to dissolution of the self–other boundary — a computational reading of one of the most striking features of psychedelic experience.

8/ Testable predictions via iterative-optimisation psychophysics: H1: G-constraint shifts reality judgements H2: C-layer choice changes complexity (low → geometric, high → object-like) H3: D-threshold tracks individual differences (e.g. VVIQ)

7/ Why it matters: DeepDream + VR can induce psychedelic-like experience without drugs (Suzuki 2017). DCNN + DGN can model neurodegenerative hallucinations (Suzuki 2024). Now: a unified architecture for why each type differs.

200

6/ This reformulates Huxley's "reducing valve." What's on the other side isn't "Mind at Large" — it's the effective causes the classifier learned, normally hidden behind G's natural image constraint. Valve opens → C's features flood into perception.

5/ Each component → a distinct hallucination type: C-type (psychedelic): G recedes, C's effective causes exposed → geometric patterns G-type (PD/DLB): veridical but wrong-category → realistic figures D-type (schizophrenia/AVH): source misattribution

310

4/ Enter the C × G × D framework: 🔵 Classifier (C): extracts features from sensory input 🟢 Generator (G): synthesises internal images (natural image prior) 🟠 Discriminator (D): judges external vs internal origin (Perceptual Reality Monitoring)

210

3/ Computational phenomenology (CP) bridges 1st-person experience and computational models. But current CP leans heavily on the Free Energy Principle. I argue we should treat the architecture of deep learning itself as a phenomenological descriptor.

320

2/ The puzzle: psychedelic geometric patterns, Parkinson's veridical visions, and schizophrenia voices are all called "hallucinations" — but they're qualitatively very different. What varies, and along which dimensions? Existing theories don't say.