We're using the wrong language to describe what ChatGPT does. It can't lie to you because lying requires intent to deceive, and the software has no intent. The

We're using the wrong language to describe what ChatGPT does. It can't lie to you because lying requires intent to deceive, and the software has no intent. The software does one thing: It generates text based on a prompt. All this other stuff that's being layered atop it, including filters to add political bias and filters to catch some derogatory speech, are after it does its thing, or before it's allowed to digest the prompt. Even without those fingers on the scale for good or ill, all ChatGPT can do is generate text from the prompt. It doesn't "know" things. It cannot process meaning. It can only produce likely text given its training text. The problem is in our anthropomorphizing the tool. It is a tool, and people who regularly use tools know their limitations and appropriate applications. We're making the tool available to a general public that has no idea what the limitations are. The public is told this is "AI" (it isn't), and they're thinking C3PO not automated MadLibs. The real thing this is going to do is drastically increase the noise, drowning out signal in our day to day lives. Nothing will be trustworthy information, because all of it will get washed through these generators.