I just generated a C# code snippet with ChatGPT, converting an integer to a base32 string. On a glance, it looked correct. But it could impossibly have generate

I just generated a C# code snippet with ChatGPT, converting an integer to a base32 string. On a glance, it looked correct. But it could impossibly have generated the given answer because it helpfully claimed how it used Crockford's Base32 algorithm (that avoids I, L, 0 and U) and had correctly adapted the code for that. However, the answer had an "L" in it. So it could impossibly be correct. And indeed, the entire answer was bogus. And besides, it forgot to tell that this algorithm also avoided "U" in its explanation of what characters it avoided. ChatGPT is littered with these things once you actually get down to business with it. This raises many philosophical questions for me. Is the AI hype trumping responsibility? When is AI good enough? "Always check your sources", sure, but the point with AI is that it can, and will, reply to questions where sources are missing by compiling a very large sum of knowledge and drawing conclusions. Do you think these are words Sam will have to eat up? Of course, Google search results can also lie, but at least you know which kinds of sites you are looking at and if they are renowned. The new Bing Chat AI _will_ list its sources but it's still drawing conclusions and compiling facts from cherry picked data from the sites. So you still end up squinting, second guessing it. I'm sure I will jump on to the new Bing AI ASAP, hell I'm a computer engineer after all and get hyped about these things, but I also admit that the failure of LLM/GPT AI to admit uncertainty, but conversely also be very willing to be bold. A dangerous combination if there ever was one that I think is an early red flag we should heed to.