Fact-checking
Generative AI can produce inaccurate or misleading information. As you read earlier in the e-course, a language model lacks consciousness and simply predicts words. It does not actually know whether the information is correct. This is why it is essential to always check the source. Compare multiple sources and evaluate them carefully.
The free version of certain language models, including ChatGPT, does not provide sources, so you cannot be sure if the output is factual or made up. When you ask the model to verify information, it can even invent fictional sources. However, Copilot and ChatGPT Plus, the paid version, do have internet access and can display sources. Even then, it is important to remain critical of both the sources and the output of the language model.
Grok, the chatbot from X, formerly Twitter, wrongly accused NBA player Klay Thompson of vandalism after misinterpreting a tweet about basketball. The chatbot claimed that Thompson was destroying homes in the US, when in reality he was talking about ‘shooting bricks’, basketball slang for missing a shot. So, never blindly trust what an AI chatbot says.
Stay alert, because despite warnings, AI systems can still provide misleading information. Take, for example, the viral photo of Pope Francis in a striking white Balenciaga jacket, which turned out to be fake. Or the AI-generated images of the Eiffel Tower on fire, as seen in the video on the right. Such fake images and misinformation can spread rapidly, especially if people do not critically assess what they see or read.
If you look closely at the Balenciaga Pope image, you can spot clear signs of AI manipulation. For instance, the crucifix on his chest seems to mysteriously float: half of the chain is missing. His right hand appears to hold a coffee cup, but his fingers grasp at nothing. Additionally, his right eyelid blends into the frame of his glasses, and the shadow of the glasses does not align correctly.
Quiz time! Can you spot an AI-generated image? Pay close attention to the details and use the tips provided with each answer. Remember this above all: AI image generators are constantly improving, so the signs you rely on today may become less obvious or even disappear in the future. Best of luck!