New Chatbot FátimaGPT Uses AI to Provide Reliable Answers: Aos Fatos Improves Fact-Checking with Language Learning Model
Brazilian news outlet Aos Fatos has integrated OpenAI’s ChatGPT technology into its existing journalism to create a conversational Q&A chatbot called FátimaGPT. This new chatbot aims to provide reliable answers to user queries by utilizing AI language learning models.
Aos Fatos, known for its fact-checking and disinformation investigations, initially launched a chatbot named Fátima in 2019. However, the bot had limitations as it relied on keyword searches in Aos Fatos’s archives to provide relevant URLs in response to user queries.
With the launch of OpenAI’s ChatGPT, Aos Fatos saw an opportunity to enhance their fact-checking capabilities. They upgraded Fátima to FátimaGPT by incorporating a language learning model with their archive database. This integration allows FátimaGPT to provide clear and concise answers to user questions, along with source lists of relevant URLs. The chatbot is available on WhatsApp, Telegram, and the web.
During the beta testing phase, FátimaGPT demonstrated impressive results. Out of the analyzed answers, 94% were deemed adequate, meaning the chatbot provided the necessary information, while 6% were labeled as insufficient. Notably, there were no factual errors in any of the responses.
Aos Fatos took precautions to address concerns about the potential spread of false information by using a technique called retrieval-augmented generation. This technique links the language learning model to Aos Fatos’s database, which contains factual information from their fact-checks and articles. When users ask questions, FátimaGPT uses this reliable database to generate answers instead of providing a list of URLs for users to browse through. This approach minimizes the workload for users and ensures the delivery of reliable information.
The use of AI technology in Brazilian newsrooms is an interesting development, particularly with regards to declining trust in the news. Aos Fatos and other media organizations in Brazil are utilizing artificial intelligence to enhance their news products and combat disinformation. A 2023 study from KPMG revealed that 86% of Brazilians consider AI to be reliable, and 56% are willing to trust the technology.
FátimaGPT has already sparked curiosity among users who are eager to test its potential biases. They ask questions about different political figures or conflicting topics to gauge the chatbot’s objectivity. Aos Fatos plans to further develop FátimaGPT to accept YouTube links, enabling the extraction of video subtitles for fact-checking against their journalism.
Efforts have been made to ensure transparency and accuracy in FátimaGPT’s responses. Users are provided with the publication dates of the shared information, allowing them to assess its relevance and potential outdatedness.
Aos Fatos has been at the forefront of utilizing AI in journalism, launching audience-facing products like Fátima and developing tools such as Escriba for audio transcription and Radar for monitoring disinformation campaigns on social media.
Other Brazilian news outlets have also embraced AI technology in innovative ways. Agência Pública uses text-to-speech technology to read stories aloud to users, while Núcleo developed Legislatech, a tool for monitoring and understanding government documents.
By leveraging AI technology, Brazilian newsrooms are demonstrating their commitment to providing accurate and reliable information in an era of declining trust in the media. Aos Fatos, with their FátimaGPT chatbot, is leading the way in combining AI with journalism to enhance fact-checking and improve the delivery of news to the public.
