Fact-Checking AI

The use of AI has become widespread in the past few years. Despite the many helpful things AI is able to do, it is still a program that runs on pattern recognition and doesn't truly understand the information it is giving. So how can someone check if the information an AI is giving you is correct? 

 

In order to properly check an AI model it is important that you check where a model is taking its sources from, and if it can filter inaccurate sources. Something I have done with AI models is use the small rock prompt. If you ask an AI model if you should eat one small rock a day, you would think it would obviously tell you not to do something like that. But, AI programs that can't properly cross check sources will cite an Onion article that is headlined “Geologist Recommends Eating one Small Rock a Day”. If an AI does not properly cross check its sources, it will assume that source is correct and tell you to eat one small rock a day while citing the Onion article 

 

It is always important to ask an AI model where it is getting the information it is giving to you. If you are asking AI a question and it gives you an answer, ask it where it got the sources for the answer it gave you. AI has a history of making up sources for its information, so if AI is not able to give you its sources, then the information likely can not be trusted. 

 

It is important to know that AI does not actually know what it is talking about. AI works by using an algorithm to analyze massive data sets, such as the internet, to make predictions to respond to your prompt. This prediction is not always the most accurate, and can take information from the internet that is often wrong, and assume that it is accurate and use that information to answer your prompt. AI has been used so much that AI generated content is commonplace on the internet. There is enough AI generated content that AI models are starting to cite previously AI generated content, which can inhibit accuracy further. AI is inevitable, and has become so ingrained in everyday society that understanding how AI works and being able to understand the risks of AI is imperative for everyone. This is so that all of us can navigate our AI-filled society more successfully.