---
title: "Informative Imitation"
description: "Beguš's term for how a speech GAN learns: the generator never sees training data; it must produce audio that fools a discriminator and encode a code (binary in fiwGAN, one-hot in ciwGAN) that a Q-network can recover from the output."
type: "glossary-term"
canonical_url: "https://antikythera.wiki/terms/informative-imitation"
md_url: "https://antikythera.wiki/md/terms/informative-imitation"
last_updated: "2026-09-06"
site: "Antikythera Wiki"
---

# Informative Imitation

> A working definition written for this wiki from the reading notes; not Antikythera's own wording. From Antikythera Wiki (https://antikythera.wiki/terms/informative-imitation), an independent guide to Antikythera's published work.

Beguš's term for how a speech GAN learns: the generator never sees training data; it must produce audio that fools a discriminator *and* encode a code (binary in fiwGAN, one-hot in ciwGAN) that a Q-network can recover from the output. The requirement to be maximally informative forces it to encode linguistically meaningful distinctions (lexical and sublexical) rather than arbitrary acoustics. The paper contrasts this with autoencoder *replication* and transformer *prediction*: imitation is a distinct learning regime, analogous to a child learning to encode meaning into imitated sounds, and is the mechanism behind imagitation and spontaneous concatenation. Whether imitation is really non-predictive (vs. the corpus's predictive-intelligence consensus) is an open tension.

## Appears in

- [Latent Spacecraft](https://antikythera.wiki/work/journal/latentspacecraft) (2026-02-04)
- [What Is Intelligence?](https://antikythera.wiki/work/book/what-is-intelligence) (2025-09-23)

_Glossary generated 2026-09-06._