Could making Wikidata 'human' readable lead to better AI?

Somewhere on Earth: The Global Tech Podcast

Could making Wikidata 'human' readable lead to better AI?

A new project is underway to allow Large Language Models (LLMs) to read Wikidata. The data is currently structured in a way that’s machine readable, but LLMs read data more like humans than machines, meaning this vast amount of human curated, high quality data isn’t accessible to this type of AI. By allowing access to Wikidata, LLMs could become more reliable. Ania spoke to Lydia Pintscher, the Portfolio Lead Product Manager at Wikidata Deutschland, to learn more about these developments.

Most news websites block AI Chatbots

Two thirds of high quality news websites block AI chatbots from accessing their information, according to a report by the misinformation monitoring organisation NewsGuard. This means that some of the world’s most popular AI chatbots could be collecting data on misinformation from low quality news sources and even conspiracy and hoax sites. The Enterprise Editor at NewsGuard is Jack Brewster and he is on the show to explain their findings.

The programme is presented by Gareth Mitchell and the studio expert is Ania Lichtarowicz.

More on this week's stories:

Wikidata and Artificial Intelligence: Simplified Access to Open Data for Open-Source Projects

AI Chatbots Are Blocked by 67% of Top News Sites, Relying Instead on Low-Quality Sources

Support the show

Editor: Ania Lichtarowicz

Production Manager: Liz Tuohy

Recording and audio editing : Lansons | Team Farner

For new episodes, subscribe wherever you get your podcasts or via this link:

https://www.buzzsprout.com/2265960/supporters/new

Follow us on all the socials:

• Join our Facebook group
• Instagram
• Twitter/X

If you like Somewhere on Earth, please rate and review it on Apple Podcasts

Contact us by email: [e-mail]

Send us a voice note: via WhatsApp: [phone]

Find a Story + Make it News = Change the World

Learn more about your ad choices. Visit megaphone.fm/adchoices