What— hieroglyphics in AI??? (and in Google?)

You’ve heard of Fabricius, right?

Actually, probably not. I hadn’t either.

Here’s Google’s description of it:

Fabricius, a Google Arts & Culture Lab Experiment that uses machine learning to help translate ancient Egyptian hieroglyphs.

That sounds implausible, but it’s true. The introductory video explains that machine learning can greatly speed up the process of translating texts that are written in hieroglyphs. Maybe so. The first thing I did was try it in Safari on my seven-year-old iMac. That didn’t work. Then I hit myself on the side of the head and said “Wait! This is a Google experiment! Let’s try Chrome!”

That worked. Even on my ancient machine.

Try it out. I do like their mix of character recognition with computerized drawing tools. But I have mixed views on their use of emojis as a teaching tool.

Why? Emojis are like hieroglyphs, right?

Well, no, actually. I can see some obvious reasons to use emojis as a teaching tool: they’re familiar to the audience, they look a lot like hieroglyphs, and sometimes they mean what they look like. But really now, most hieroglyphs are phonetic; only a small minority of the time do they represent entire words. Often a hieroglyph denotes the category of a word, but that’s not really emoji-like either.

Anyway, it’s early days yet. It’s a cool idea, so let’s wait and see. I am pleasantly reminded of my very first programming project of any significant size: in my senior year of college, I developed a program on the PDP-1 to test myself on my growing knowledge of Egyptian hieroglyphs, a task that initially involved inputting a lot of characters with the aid of a light pen. Since you probably don’t know what a light pen and a PDP-1 are, I will include a couple of photos (a modern image of a light pen, followed by an old black-and-white from the ’60s that includes a PDP-1) and then a lovely image of 4000-year-old hieroglyphics with their original colors still intact:




Categories: Linguistics, Technology