The document combines a technical perspective by Aniket (Niki) Kittur and the full research paper "Soylent: A Word Processor with a Crowd Inside" (Bernstein et al.). It argues that embedding paid crowd workers from Amazon Mechanical Turk directly into a word‑processing UI can provide high‑level editing assistance—text shortening, proofreading, and arbitrary macro tasks—that current AI cannot reliably perform. The core technical contribution is the Find‑Fix‑Verify programming pattern, which decomposes open‑ended editing tasks into three stages (identifying problematic patches, generating edits, and voting to verify quality) to mitigate the high error rates typical of crowd work. Soylent implements three user‑visible components: Shortn (interactive paragraph shortening via a slider that selects among crowd‑generated rewrite combinations), Crowdproof (human‑augmented spell/grammar/style checking with drop‑down suggestions), and the Human Macro (natural‑language task specification for arbitrary document manipulations). The system is built as a Microsoft Word add‑in (VSTO/WPF) with a TurKit‑driven backend. Evaluation across five text corpora shows Shortn reduces length to 78‑90 % of the original with median wait times of ~18 min (mostly recruitment latency) and median work time under 2 min; Crowdproof catches 67 % of manually labeled errors (vs. 30 % by Word) and fixes 88 % of those; Human Macro achieves 88 % intention success and 71 % accuracy. The paper discusses limitations (latency, cost, privacy, legal ownership, domain knowledge) and broader societal implications of crowd‑powered interfaces, urging interdisciplinary research on incentive design, reputation, and ethical labor practices. The perspective piece expands on these themes, positioning crowd‑powered systems as a new class of interactive software that could eventually replace “Wizard‑of‑Oz” prototypes with permanent “Wizard‑of‑Turk” components, and speculates on future applications in creative domains.
Topics
Vision of Permanent Crowd‑Powered Interfaces
Kittur’s argument that embedding paid crowd workers into interactive systems can create new forms of computational support beyond current AI capabilities.