What ChatGPT Knows About Exploring Binary (Spoiler: Not Much)

Like everyone else these days I’ve been playing around with ChatGPT:

It can “write a biblical verse in the style of the King James Bible explaining how to remove a peanut butter sandwich from a VCR” but it’s never heard of my site.

(“Useful resource” is flattering, especially after first hearing it did not know about my site.)

A bit generic but not bad. I do cover the basics of binary numbers, but what makes this blog unique is the discussion of the intricacies of decimal to binary floating-point and binary floating-point to decimal conversions, including the related bugs I’ve found in commercial and open-source software.

Again, pretty generic, and not the specifics I was looking for. My converters and calculators are my most popular pages, followed by the articles Number of Bits in a Decimal Integer, Binary Multiplication, Binary Division, Ten Ways to Check if an Integer Is a Power Of Two in C, and Binary Subtraction. (The “intricacies” articles have a much smaller potential audience.)

(“Clear explanations”. Wait – does it know or is it just continuing to try to flatter me?)

Some of the links I see in my stats from US colleges and universities are from Drexel University, Emory University, New York University, Stanford University, Stony Brook University, University of Texas, University of Washington, Villanova University, and Virginia Tech.

“Bicimal” was a term I found in the bowels of the internet when I was looking years ago for a way to describe fractional binary numbers. I wrote a few articles using it, and the top search results on the term lead to this site.

Another generic answer, although I certainly agree with “make it fun”. I prefer my method though, using tape flags (analogous to base ten blocks) and describing another non-decimal base (base 5) before getting to base 2 (which in its simplicity almost hides the place value pattern).

That’s basically the same answer as for third graders. I found that a different method might be more suited to adults, approaching it by showing that any number can be written as a sum of powers of two (and not calling them “powers”).

“Why does 0.3 + 0.6 = 0.89999999999999991?” was a sentence in my article Why 0.1 Does Not Exist In Floating-Point. It got the answer right at a high level (and with some extra fluff). But I was looking for a more concise answer like “because most decimals have infinite representations in binary, and binary-floating point has finite precision”.

Pretty close, but wrong. It’s actually the largest subnormal number, not the smallest normal number.

I was looking for PHP Hangs On Numeric Value 2.2250738585072011e-308, or even its offshoot, Java Hangs When Converting 2.2250738585072012e-308. (Those two articles went viral, inasmuch as these kinds of articles do.)

That one hurt. It knows about specific websites, just not mine. It said earlier that my site dealt with two’s complement (a generic guess?), yet it missed my Decimal/Two’s Complement Converter, which has been the top Google search result for many years. To boot, none of those sites listed has a two’s complement converter (as far as I can tell).

I really had no expectations for this one, especially at this point. Overly generic, and bullet 5 is wrong.

I would have bet it would have at least picked one of the many Rick Regans that aren’t me.

“Clear and concise”. That’s three times now; I’ll accept the compliments.