Have you ever really taken some time to ask yourself what makes something true? Is something true because there are objective truths it corresponds to? Is it true if everyone agrees upon it? Is it true if believing it leads to successful results? These are deeply human questions and they impact everything we do in our lives even if we never think about them, and as humans we get to decide for ourselves what foundational truths we use to make our decisions, AI isn’t so lucky.
Every day we’re being bombarded with news about AI or fear that AI is going to take our jobs, or someone trying to sell us their online AI course, but beyond understanding how to use AI (which is absolutely vital for success) it is equally important to understand our uniquely human position and skills that will allow us to do things AI is fundamentally incapable of doing. One of those abilities is freedom to determine what is true for ourselves.
Without going into the nitty gritty of it: AI in the sense most people talk about is a “Large Language Model” which is made up of artificial neurons all trained on (ideally) human created content, then from that content and training it can essentially guess what the right words are to provide to you, the user, based on a combination of your question and it’s knowledge, often with some search engine results thrown in as well. That is incredibly effective, but how does it know that what it’s providing to you is fundamentally true?
AI isn’t truly conscious in the same way we humans are, so it relies on its programming and therefore is at the mercy of the consensus of its training data. As such AI must, by its very design, operate by what is sometimes called the “consensus theory of truth” which essentially says something is true if most people would agree that it is true. See the problem here? That theory of truth is useful if and only if most people are right. There are things engineers do to try and combat this by adding trusted sources but that just moves the problem up one level, now something is true to AI if most “experts” (as defined by the engineers) would agree it is true. Again, this can be very practical but if experts get something wrong, which any expert in any field will tell you happens often, AI is fundamentally incapable of having that human intuition to see it doesn’t line up, or to push for more, or to take the less universally agreed upon answer because we know deep down inside that it’s true.
As AI becomes a greater and greater part of our lives a huge part of remaining able to provide value to the world in a practical sense is going to be tapping into our unique human superpowers, and the ability to think in truly novel ways, throwing aside what the herd believes will become more and more valuable. So what’s the practical takeaway from this? Don’t rely on AI too much to tell you what’s true and really truly understand its limitations so you can get better than a “C” in life.