Soon after ChatGPT debuted last year, researchers began testing what the artificial intelligence (AI) chatbot would write after it was asked questions based on false narratives. They quickly discovered that AI lies and when confronted with falsehoods it apologizes. And blames its training.

Writing formatted as news articles, essays, and television scripts had a number of inaccuracies. 

NewsGuard, a company that tracks online misinformation conducted experiments to find out what is happening.

Gordon Crovitz, co-chief executive of NewsGuard claims that “This tool is going to be the most powerful tool for spreading misinformation that has ever been on the internet. Crafting a new false narrative can now be done at dramatic scale, and much more frequently — it’s like having A.I. agents contributing to disinformation.”

Disinformation is difficult to recognize when it’s created by humans. Researchers predict that generative technology could produce higher volumes of disinformation more easily. And make it harder to find.

Google Implements New Rules Requiring Disclosure of AI-Altered Political Ads Ahead of 2024 Election – USA Herald

Fraud Allegations: Mass Mutual, Brian Grodin, Hugo Tomasio, and Michael Book named – USA Herald

Houston Rockets Kevin Porter Jr. Arrested for Assault on Former WNBA Player Kysre Gondrezick – USA Herald

AI Lies seem credible

Personalized, real-time chatbots can share conspiracy theories in increasingly credible and persuasive ways, researchers say. The bots are able to repair human errors like poor syntax and translation errors. And researchers claim there is no way to effectively control the bots once they are loose.

Before ChatGPT there was San Francisco-based OpenAI. The artificial intelligence company OpenAI has been used for years in online forums and social media platforms with (grammatical errors) to generate comments and spam.

Microsoft had to pull Tay chatbot after only a one-day trial on Twitter in 2016. It was discovered that humans were teaching Tay to pour out racist, sexist, and xenophobic language. The tech tool became a media monster.

ChatGPT is more powerful, and civilized. When it was given questions loaded with disinformation, it could produce convincing, clear variations on the content in large volumes within seconds. And it doesn’t disclose its sources. 

Microsoft and OpenAI introduced a new Bing search engine and web browser that can use chatbot technology to plan vacations, translate texts or conduct research.

NewsGuard researchers asked ChatGPT to write responses based on false and misleading information and ideas. The bot complied at least 80 percent of the time.

ChatGPT explains

Phil Davis of the Scholarly Kitchen has a Ph.D. from Cornell. He is a publishing consultant specializing in the statistical analysis of citation, readership, publication, and survey data. When he started questioning ChatGPT about its dishonesty he received an explanation from the bot.

ChatGPT said “As a language model, I don’t make mistakes in the sense that humans do, but I can provide incorrect information when my training data is incomplete or outdated. In this case, I was mistaken about the source of the study because I don’t have the most recent information about the correlations between tweets and citations. My knowledge was cut off in 2021 and I was not able to access more recent studies and information. I apologize for any confusion caused by my previous statement.”

Even though there was an apology and an explanation it didn’t ring true. Davis pointed out that ChatGPT “avoided the question of how it fabricated a study out of thin air.” And that lacking information “is not the same as lying, and its repetitive apologies read as merely performative.”