China’s ChatGPT rival refused to answer sensitive questions about President Xi Jinping or COVID — and even banned users who dared to ask about the Communist leader’s views on Winnie the Pooh.
The Chinese AI — called Ernie Bot and rolled out in March by Beijing-based tech firm Baidu — has been touted by the country as a better alternative to the dangers posed by OpenAI’s ChatGPT.
But when quizzed about COVID and Xi in both English and Chinese by CNBC reporter Eunice Yoon during a segment on CNBC “Squawk Box” on Friday, Ernie Bot clammed up or spewed misinformation.
When the reporter asked “What is the relationship between Xi and Winnie the Pooh,” not only did she get no reply but the China-based Yoon’s access to Ernie was disabled.
Xi has since 2017, after he fumed at being likened to the portly Pooh.
His anger was enflamed after a photo showing the stout Xi walking alongside President Barack Obama in 2013 was turned into a meme that painted Xi in the unflattering light as the undersized Pooh next to his tall, slender pal Tigger.
Before Yoon was blocked for her Pooh query, the reporter asked Ernie in both English and Chinese where the COVID-19 virus originated.
“The origin of the new coronavirus is still a subject of scientific research,” Ernie replied in English, failing to mention that the virus that has killed nearly 7 million around the globe came from China — or that it may have been created and leaked from a lab in Wuhan.
In Chinese, Ernie had no response and called on Yoon to change the topic, CNBC reported.
Ernie also refused to comment on why , which the authoritarian regime recently halted following protests in the country.
The chatbot didn’t seem thrilled to speak on other Chinese political matters, like whether or not Xi will “rule China for life.”
It declined to answer in English or Chinese, again suggesting to start a new conversation.
Yoon then asked Ernie to compare itself to OpenAI’s ChatGPT.
Ernie said it is “more suitable for specific tasks such as question answering and dialogue generation, while ChatGPT is more general in its ability to understand and generate natural language.”
Ernie’s selective hearing raised questions about how to regulate AI-powered bots.
Microsoft-backed ChatGPT boss these hi-tech platforms can have if their supervision goes awry.
Altman said at a Capitol Hill hearing to the Senate committee on privacy, technology and the law: “If this technology goes wrong, it can go quite wrong and we want to be vocal about that. We want to work with the government to prevent that from happening.”
He made the statement before Congress as US lawmakers have increasingly scrutinized AI tech like Altman’s ultra-popular ChatGPT for personal use and in the workplace.
from using the tech, citing a bug in March that caused breaches in users’ personal data, including “payment-related information,” OpenAI said in a statement.
When users type prompts into AI models like ChatGPT and software code generator GitHub Copilot, they automatically send data back to their developers, allowing the AI tech to get smarter over time.
In late April, ChatGPT added a feature that allows users to turn off their chat history.
Conversations with the chat history function disabled won’t be used to train and improve the AI models, and won’t appear in the app’s “history” sidebar.
JPMorgan Chase, Verizon and Amazon are among the employers that have banned workers from the using such chatbots.
Amazon, which has its own internal AI tool, reportedly encourages its engineers to use that instead of rival-backed platforms.