Content
ChatGPT through the lens of the Dunning-Kruger effect! 100 1 Wow. ChatGPT is able to answer all my questions. It quickly generates very elaborated answers that are credible and coherent. Finally a chatbot that behaves like a super human. Ready now. Now I understand where the real value of ChatGPT resides. It can definitely give my productivity a boost, but it’s important to recognize when it gets things completely wrong. HFS Confidence level in ChatGPT Wait a minute. ChatGPT is powered by a Large Language Model (LLM) which is ultimately a statistical tool used to predict language without understanding it and produce "statistically plausible" answers. 0 Damn. This means that ChatGPT will occasionally generate incorrect answers by unintentionally stitching wrong snippet of information together. Got it. ChatGPT is great when there isn’t a precise "right answer". But it cannot be trusted when the answer must be "perfect" to be reliably useful. No knowledge (believe the hype) Knowledge of ChatGPT Enough knowledge (understand the reality)