There, I said it. AI has proven to be an interesting and sometimes useful tool for me. But honestly, at the moment, it is little more than an aggregator of information and an articulator of answers.
Step One: You ask AI a question.
Step Two: AI plows through an array of publicly available sources.
Step Three: AI molds a bunch of information into an understandable (but not necessarily meaningful or accurate) response.
And while we’d like to close our eyes and pretend most of the sources AI uses are scholarly papers and vetted news stories providing researched, factual data, that would be a huge mistake. Because many (and maybe even most) of those sources suck, as does AI’s ability to know the difference.
According to AI (in this case, Perplexity), “The inability of AI to consistently and reliably distinguish between high-quality and questionable data or sources is indeed a major limitation, affecting accuracy, trust, and usefulness of outputs.”
In other words, its limitations are built into the system. AI models rely heavily on their training data, which often includes vast amounts of uncurated, biased or outdated material scraped from the internet. Because most models are designed for pattern recognition rather than truth verification, they can reproduce errors, amplify biases and propagate misinformation present in their sources.
To be clear, AI is a potentially great tool, assuming you understand and compensate for its weaknesses. That is to say, DTAI (Don’t Trust AI); at least don’t trust that the answers you get to every prompt are actually good answers.
Remember the principle of GIGO (Garbage In, Garbage Out)? Basically, it states: if you put incorrect or low-quality data into a computer system, you’ll get incorrect or low-quality results, regardless of how sophisticated the system is. Same with AI.
And AI readily acknowledges that it cannot distinguish between high- and low-quality sources. Which means WE have to be diligent when using AI. Truth verification is our responsibility. In short, the problem isn’t AI, it’s US.
