AKASHICOMNI × AKASHICNET · CONCEPT REVIEW · 26 SEPTEMBER 2026

A conversation with Dennett’s ideas.

Bring a text. Test an intuition. Pour a glass of wine. Keep the chair empty.

AI-generated concept art · no likeness, encounter or working chat shown

THE QUESTION

Could we talk with a philosopher’s work while remembering that the philosopher is absent?

Imagine asking how Dennett’s published arguments bear on a new essay about consciousness, AI or free will. A useful system could find passages, compare positions and push back with a thought experiment. It would have to show which part comes from a source and which part is its own inference.

And the drink? Tufts recounts that Dennett made apple brandy and blueberry wine on his Maine farm. The illustrated wine glass nods to the invitation to talk; it represents no real meeting.

ESTABLISHED RESEARCH

Someone already tried the imitation.

Researchers fine-tuned GPT-3 on Dennett’s writings, asked both Dennett and the model ten philosophical questions, and asked 425 participants to find his actual answers among four generated alternatives. Experts on Dennett’s work chose the real answer 51% of the time, compared with a 20% chance baseline.

That is evidence that generated answers could be difficult to distinguish in this task. It does not establish that the model understood his arguments, held his views or could predict what he would say about a new text.

Read the study ↗

THE ETHICAL COUNTERWEIGHT

The person must never be counterfeited.

Dennett later warned about AI-generated “counterfeit people” that pass for real people online. His essay does not evaluate this proposed educational tool. Its warning gives us a clear design test: the interface must never make readers wonder whether Dennett is actually present, endorsing an answer or replying from beyond his lifetime.

Read Dennett’s essay ↗

PROPOSED EXPERIENCE · INTERPRETATION

Three modes. Three different claims.

01 · SOURCE

What Dennett wrote

A concise account with a verifiable work, passage and date. Quotation remains limited by rights.

Label: established source

02 · INFERENCE

Apply the framework

Explore how published concepts might illuminate a new text, including tensions and alternative readings.

Label: AI-assisted interpretation

03 · FICTION

At the pub

A playful exchange about an idea. Its imagined voice is clearly marked throughout and never attributed to Dennett.

Label: fictional simulation

IMPLEMENTATION PATH · PROPOSAL

How to do it responsibly

Inventory the works. Record bibliographic details, rights holder, access conditions and whether each full text may be indexed or quoted.
Start with retrieval. Search rights-cleared material and cite relevant passages before generating an answer. Do not claim to have trained on “all of Dennett.”
Separate answer modes. Distinguish source summaries, new analysis and imagined dialogue in both wording and visual treatment.
Evaluate with readers. Test fabricated citations, overconfident predictions, conflicting passages, requests for long excerpts and posthumous attribution.
Publish the audit trail. Document corpus coverage, exclusions, model and retrieval versions, corrections, costs, privacy and rights decisions before opening a public chat.

AKASHICOMNI LENS · EXPLORATORY

When does a model of ideas become a misleading model of a person?

AkashicNET could hold the source map, rights ledger and corrections. AkashicOMNI could compare the philosophical, social and technical readings, then test whether people understand the labels. The empty chair is a design reminder, not evidence for any claim about consciousness or survival after death.

FULL TRANSPARENCY REPORT · VERSION 1 · 26 SEPTEMBER 2026

What this page can and cannot claim

Status: Editorial concept and generated artwork. No Dennett corpus was assembled here; no model was trained, chatbot deployed or user study conducted.
Research basis: One direct model-comparison paper; Dennett’s own essay; a Tufts archive catalogue and profile; and US Copyright Office reports. The paper’s reported percentages refer to its specific test, not this proposal.
Editorial interpretation: The three modes, empty-chair metaphor and proposed safeguards are AkashicOMNI × AkashicNET design choices. They are not statements by Dennett.
Source and rights boundary: The catalogue is a lead, not a licence. Individual publishers and rights holders may control books, essays, interviews, images and recordings. Rights review is pending for any actual corpus.
Visual disclosure: The hero is AI-generated from a written brief. It depicts no portrait, genuine pub meeting, product screen, source corpus or measured scientific network.
Uncertainty: Corpus completeness, permission, cost, suitable evaluation, estate preferences and public-launch suitability remain unresolved. No endorsement by the estate, publishers, Tufts or the researchers is implied.
Correction rule: A future product should expose source-level corrections and version history. Generated first-person speech must never be presented as an authentic Dennett quotation.

SOURCE TRAIL

Read the material behind the concept

Creating a Large Language Model of a Philosopher ↗Eric Schwitzgebel, David Schwitzgebel & Anna Strasser · research paper

The direct precedent: a model fine-tuned on Dennett’s writing and compared with his own answers.

The Problem With Counterfeit People ↗Daniel C. Dennett · The Atlantic · 16 May 2023

His warning about systems that pass themselves off as people.

Daniel C. Dennett papers ↗Tufts Digital Library · archive catalogue

A potential source inventory; catalogue access does not settle reuse rights.

Our Brains, Our Selves: Daniel Dennett ↗Tufts Now · profile

Tufts recounts his Maine farm, apple brandy and blueberry wine.

Copyright and Artificial Intelligence ↗US Copyright Office · reports

Separate reports address digital replicas and generative AI training.