META created AI to verify the facts in Wikipedia

META has developed an artificial intelligence algorithm, automatically checking footnotes in Wikipedia. He can also offer alternatives if the passage is inaccurate.

(1/2) Today, We’re Announcing The First Model Capable of Automatically Verifying Hundreds of Thousands of Citations at Once. Read More: https: // t.CO/C0NNVHUQX7 PIC.Twitter.COM/UTSEYUMCRZ

– Meta ai (@metaai) July 11, 2022

Sphere uses a model of transforming an understanding of a natural language (NLU), which is trying to understand the different relationships of words and phrases in a sentence.

SPhere database consists of 134 million web pages. The algorithm for understanding the natural language analyzes footnotes in the articles of the encyclopedia and is looking for a single source to verify each statement.

According to company representatives, Sphere can analyze “tens of thousands” of quotes at the same time. META noted that the electronic encyclopedia is replenished with 17,000 articles at the same time containing many footnotes that are difficult to check manually.

To illustrate the capabilities of AI, Meta https://gagarin.news/news/pay-anywhere-pay-easily-pay-safely-ecb-unveils-digital-euro-motto/ showed an example of an incomplete quote found by the model on the Wikipedia page about the indigenous American people of black -legged people. The section “Famous Persons” (in the English -language version of the article) is mentioned by Joe Hipp – the first representative of the nation who fought for the title of world champion according to WBA in heavyweight.

However, on a connected web site, it was not possible to find confirmation of this fact. Having searched in the SPhere database, the model revealed a more suitable quote in the 2015 article at Great Falls Tribune. The algorithm noted the following excerpt:

“In 1989, at the sunset of his career, [Marvin] Kamel fought with Joe Hippes from the nation of black -legged. Hippe, who became the first indigenous American to challenge the world of heavyweight world, said that this battle was one of the strangest in his career ”.

The box found is not directly mentioned in the passage. Sphere discovered a suitable link due to the possibilities of a natural language, the developers emphasized.

The company believes that in the future, the tool will help to fight misinformation on Facebook.

“In a more general way, we hope that our work can be used to verify the facts and increase the overall reliability of information on the Internet,” the creators of the model said.

In addition, META plans to create a platform that Wikipedia editors will be able to use for systematic verification and correction of footnotes.

The source code of the model is available on github.

Recall that in July, Meta was presented by AI-translator who supports 200 languages.

In April, technical gigant spoke about the development of the OPT-175B language model with open source code.

Subscribe to FORKLOG news at Telegram: Forklog AI – all news from the world of AI!