Meta Launches An AI knowledge Tool Called “Sphere” Based On Open Web Content

Facebook says it wants to help fight misinformation rampant across the internet in recent years. But it’s an issue it may have helped create in the first place. Facebook parent company Meta announced a new AI-powered tool on July 11 called Sphere. The Meta Sphere AI is intended to help detect and address misinformation, or “fake news,” on the internet. The company claims it is the first AI model that can automatically scan numerous citations at once.

How It Works 

The idea behind applying Meta Sphere AI for Wikipedia is quite simple. The online encyclopedia contains 6.5 million entries. On average, it displays almost 17,000 articles added monthly. The wiki concept behind that means that adding and editing content is crowd-sourced. While there is an editorial team, it is still a tiring task that increases daily.  

At the same time, the Wikimedia Foundation oversees Wikipedia and has been weighing up new ways of leveraging all that data. 

Meta’s announcement regarding a partnership with Wikipedia does not reference Wikimedia Enterprise. Adding more tools for Wikipedia ensures that its content is verified and accurate. So, it will be something that target Enterprise service customers will want to know when considering paying for the service.

Meta has confirmed that there is no financial arrangement in this deal. However, the company notes that training the Meta Sphere AI model created “a new data set (WAFER) of 4 million Wikipedia citations.

What Can Happen in The Future

Based on the comments that TechCrunch got from Meta, here are the things that might happen soon. 

• Meta believes that the Sphere’s “white box” knowledge base has significantly more data than traditional “black box” knowledge sources. The 134 million documents at Meta used together to develop the Sphere AI were split into 906 million passages of 100 tokens each.

• By open sourcing this tool, Meta argues it’s a more solid foundation for AI training models and other work than any proprietary base. It concedes that the very foundations of the knowledge are potentially shaky, especially in these early days. What if a “truth” is not reported as widely as fake news? That’s where Meta wants to focus its future efforts in Sphere. Meta intends to develop models to evaluate the quality of retrieved documents, detect potential errors and prioritize more credible sources. 

• Meta Sphere AIl raises some interesting questions on what Sphere’s hierarchy of truth will be based on compared to other knowledge bases. Because it’s open source, users may be able to tweak those algorithms in ways better suited to their own needs.

• Meanwhile, Meta has confirmed that it is not using Sphere or a version of it on its platforms like Facebook, Instagram, and Messenger, which have long grappled with misinformation and toxicity from bad actors. (We have also asked whether other customers are in line for Sphere.) It has separate tools to manage its content and moderate it.

• The development seems to be designed for mega scale. Wikipedia’s current size has exceeded what any great team of humans could check for accuracy. The critical point is that Sphere is being used to automatically scan hundreds of thousands of citations simultaneously to spot when a citation doesn’t have much support across the wider web. 

While Meta Sphere AI is still in the production phase, it also seems like the editors might be selecting the content, which might need verifying for now. Eventually, Meta’s goal is to build a platform to help Wikipedia editors systematically spot citation issues and quickly fix them or correct the content of the corresponding article at scale.

And for other news stories, read more here at Owner’s Mag!