AI lies to you in 3 different ways. Hallucination = making it up. Sycophancy = sucking up. Bias = lopsided answer. #aitips #aiforbeginners #aiworkflow #claudeai #aihacks

AI lies to you in three different ways and most people only know one of them. It does it through hallucination, sycophancy, or bias. And all three are just AI getting things wrong in ways that usually look right. Hallucination is when AI makes up facts that seem completely real. So this would be if Claude gives you a stat with full confidence, and the source doesn't even exist. The fix is just one line that you can add to all your prompts. Say don't ever guess. If you're not 100% sure that something is true, say I don't know and go find the answer instead of ever giving me false information. Sycophancy is when AI agrees with you when it probably shouldn't. So let's say that you share a crappy business idea that would never actually succeed, and AI tells you that it's absolutely brilliant. The fix again is just one line that you can add to any of your prompts. Say don't tell me what I want to hear. Tell me what a smart skeptic would say. And the third one bias is when AI gives you skewed answers based on what it was trained on. So certain perspectives will win and others will be buried. It's kind of like being in a political echo chamber if you're only listening to your news from one source or one station. The fix is adding a line like this, say argue with yourself. Take the opposite position and make the strongest case that you can. This will force you to see the other side in most scenarios. The easiest way to remember it is hallucination is making it up. Sycophancy is sucking up. And bias is making incorrect assumptions.