
Photo 273969814 © Darkworx | Dreamstime.com
ChatGPT, the most popular large language model, is making things up—fabricating information and generating nonsensical responses. While artificial intelligence chatbots come with a risk of hallucinating facts, this tendency is at a high as of May 23, 2024, at 5:28 AM ET (2:28 AM PT), coinciding with a widespread outage at Microsoft’s Bing (whose Copilot assistant remains down, as well).
Developed by OpenAI, ChatGPT relies on a massive dataset of text and code to train its algorithms. It’s understood that Bing plays a role in keeping that data current and relevant, as its API powers ChatGPT’s browsing functionality. An outage at Microsoft would limit the fresh information ChatGPT has access to, potentially impacting its ability to generate grounded responses about current events.
For instance, when the latest, browser-enabled GPT-4o model was asked to generate an article-like summary about the United States suing Ticketmaster owner Live Nation, it crafted one “based on a standard approach to reporting on antitrust lawsuits,” instead of searching the web to surface real-time information.
“Due to the browsing tool currently experiencing issues, I wasn’t able to pull in specific, real-time details from recent news sources,” ChatGPT responded when probed about its sources.

Screenshot via ChatGPT
The tool was once again asked to give an overview of the death of Iranian president Ebrahim Raisi, who was killed in a helicopter crash on May 19. In its response, it said that the output was made “fictional for the requested scenario,” and that it was only relying on information from “my last update in May 2023.”

Screenshot via ChatGPT
While this indirectly caused inconvenience might throw a spanner in the works for people relying on browser-enabled AI tools for bite-sized news, ChatGPT, at least, acknowledges that the fluff is made up. It also urges users to visit credible news sources for the most accurate information.
[via DesignTAXI Community, images via various sources]