Omnipedia is an English-language online encyclopedia founded by internet entrepreneur Tony Hsiung in June 2049. Intended as “a general reference work and trusted source of information,” Hsiung initially conceived of Omnipedia as a successor to Wikipedia, which was discontinued in 2048.
Publicly launched on September 28th 2049, Omnipedia currently has
26,72027,337 articles in different stages of development. It is edited by a select community of human and artificial intelligence (AI) contributors, with Hsiung as editor-in-chief.
Although Omnipedia’s pilot programme was made possible through an investment by Zhupao Campus, its public version will be funded primarily through donations in a transparent process to avoid any appearance of advocacy.
Omnipedia was officially announced on June 18th 2049 by internet entrepreneur Tony Hsiung as part of a new round of investments from Zhupao Campus. In a livestream with Zhupao executive chair Xu Shaoyong, Hsiung explained Omnipedia’s name as a reference to “the encyclopedia of everything.” When asked whether it was intended as a replacement of Wikipedia, Hsiung described Omnipedia as “more of an unofficial successor, a reboot instead of a sequel” and said that it would “follow the spirit, if not the letter of Wikipedia, which was truly irreplaceable.” 
Omnipedia went live on September 28th 2049, with Hsiung and Xu taking part in a press conference to celebrate its public launch. They described Omnipedia’s mission statement as “offering an easily accessible primary source of balanced, responsible, and accurate information about anything and everything.” Although it was initially hinted that some of Wikipedia’s content would be carried over to Omnipedia for its initial launch, Hsiung stated that this would be infeasible due to the copyright issues that had contributed to Wikipedia ceasing operations in April 2048. As a result, Omnipedia will feature fully original content. 
Following the death of Xu on October 1st 2049, Hsiung appeared on a personal stream to offer his condolences and announce that Omnipedia would serve as “an inclusive source of breaking news.” He also apologised for the readability and matching issues with Omnipedia’s revision histories, and promised to “greatly speed up the process of properly citing sources.”
Omnipedia articles are created in a formal two-step process. A first draft of each article is assembled by a set of proprietary differentiable neural computers (DNCs) based on the PACOTTI architecture. Following prescribed research and collation algorithms, the DNCs gather information from a wide variety of sources to ensure a multitude of opinions and viewpoints. They structure their findings in a “comprehensive and comprehensible” text with an eye for a “well-written narrative,” list up the accompanying citations, and add complementary visual material. According to Hsiung, the Omnipedia DNCs are designed to “convey the highest and most succinct tier of any pyramidal construct of knowledge.”
The second step involves a quality assurance review carried out by human editors, who are recognised experts in their respective subject fields. Using a Drupal-based editing system, the human editors polish and refine the articles to improve readability and make sure they adhere to a consistent house style.
At present, most internal Omnipedia hyperlinks are limited to hover-only previews of their pending articles. One of Omnipedia’s operating algorithms tracks which previews are opened more than others and assigns the corresponding topics a priority weighting value for development. No announcements have been made regarding Omnipedia editions in other languages, but Hsuing has stated that these will be forthcoming after the English-language version is established. 
On October 10th 2049, The Guardian published an op-ed claiming that Omnipedia is working in the opposite direction of its intent as “a tightly curated mouthpiece for corporations and sponsors.”  Based on testimonies from an anonymous group of Omnipedia editors, the op-ed describes how Omnipedia’s use of both human and AI editors resulted in “an inadvertent correction for the distorted perception of reality propagated by digitally generated content due to model collapse.” 
In this way, Omnipedia may have contributed to “the general public being made aware for the first time of the magnitude of what happened during the CMD pandemic and the decade after, a stretch of time that has been clouded by rampant disinformation.” The op-ed also singled out Hsiung’s perceived role in allowing Omnipedia to “accidentally tell the truth, either through intent or ineptitude.”
- Hongli, P. (June 2049). “Zhupao announces Omnipedia project as ‘successor’ to defunct Wikipedia.” South China Morning Post. ↩
- Hongli, P. (September 2049). “Xu Shaoyong officially cuts ribbon on Zhupao-backed Omnipedia.” South China Morning Post. ↩
- Leyba, R. (September 2049). “How Omnipedia is put together.” Wired. ↩
- Passaic, H. (October 2049). “Omnipedia may be doing the world a favour by accidentally telling the truth.” The Guardian. ↩
- Shumailov, I; Shumaylov, Z; Zhao, Y et al. (May 2023). “The Curse of Recursion: Training on Generated Data Makes Models Forget.” arXiv. ↩