Breaking News

AI Maybe Revives Dead Languages


While Star Trek’s transporter is hard to imagine — perfect matter movement across vast distances with no equipment on one end — it may not be the most far-fetched piece of tech on the Enterprise. While there are several contenders, I strongly suspect the universal translator is the most unlikely MacGuffin. After all, how would you decipher a totally unknown language in real-time? Of course, no one wants to watch 30 episodes of TV about how we finally figured out what Klingons call clouds, so pretty much every science fiction movie has some hand-waving explanation for speaking the viewer’s language. Farscape had microbes, some aliens have telepathy that works with alien brains of any kind, and still others study English from afar for decades off camera. Babelfish anyone?

I was thinking about this because of an article I read by [Alizeh Kohari] about [Jiaming Luo’s] work using AI to decode dead languages. While this might seem to be similar to Spock’s translator, it really isn’t. Human languages change over time and distance. You only have to watch the BBC or read something written by Thomas Jefferson to see that. But there is still a lot in common, at least within certain domains.

You’re Only Human

If you are a native English speaker, you can probably puzzle out a lot of words given a text in, say, Spanish or French. Most of these languages either started with or borrowed from Latin and share roots with languages like Greek, so you can often puzzle out a sign with a little context. Now try that with Arabic or Mandarin. Most of us don’t have a clue. Do you read right to left? What are the characters? There’s nothing to grab onto unless you know the language or one that is similar. For example: “每天閱讀 Hackaday” probably doesn’t give you any clues other than it is about Hackaday.

The AI isn’t much different. The software learned how languages change as they evolve by studying patterns between related languages like Ugaritic and Hebrew. You might think languages develop ad hoc, but there are definite patterns — at least, among human languages.

For example, although not all languages have the same words for colors, it has been known since around 1969 that humans tend to develop words for colors in the same order. That is black and white get names along with red. Then, later, names for other colors appear. You can watch an interesting video about this effect below.

Close Relations

Assuming that the AI could translate one related language into another, what happens if you feed it languages that are related, but we no longer know how to read?

Seal containing Indus Script
An impression of a cylindrical seal from Iran dating back a few thousand years. Photo by [ALFGRN], CC-BY-SA-2.0.
For example, [Kohari] points out Indus script — Harappan —  found on thousands of small stamps or seals around Pakistan and as far away as Iraq. Despite decades of analysis, no one knows what they say. There have even been theories that the seals say nothing, but mathematical analysis shows the symbols are not placed randomly and appear to correlate with other languages.

The problem for the AI is that no one knows what languages might be related to the Indus script. [Luo] and his team have done work to devise an algorithm that can tell how close two scripts are even if it can’t understand them. This could help. We’ll see if his computer can eventually read the dead Indus language.

If this works, it could open up a lot of archeology. There are many languages that have been forgotten like Etruscan and Rongorongo. If you knew they had morphed into other languages we do know, this could help unlock them. Even languages we do understand fairly well today were not always known in modern times. Egyptian hieroglyphics, for example, were a mystery until the discovery of the Rosetta stone which had the same message in the hieroglyphs, Egyptian demotic script, and Greek. At least we knew how to read Greek.

AI Hype

If we had the true universal translator, it would be fun to see what the dolphins and the bees are saying to each other. It seems like if we could do that, reading Indus would be easy. The truth is, though, AI is a long way from being able to totally replace humans and human insight into creative problem solving and that’s important to remember as people look to us as people who understand technology to help them make decisions about AI.

For example, researchers think the people who wrote the Indus script seals wrote right to left because sometimes the left characters are scrunched up as sometimes happens when you get close to the edge of your paper. However, since the seals were probably meant to be used as stamps, that doesn’t necessarily imply the language itself is right to left. For instance, Chinese used to be written top to bottom and right to left, but in modern formats is written row-wise left to right. Both styles persist, but you know which way to read from context. Insights like that are still the purview of humans, at least for now.

I have read recently that AI may already be conscious. If you know much about how the brain works and how modern AI works, you’ll probably find that statement to be as unlikely as I do. We may one day replicate an electronic brain that embodies that thing that makes us “human” or even just conscious. But that day seems far away, indeed. Besides, if we do, who is to say we won’t have as much trouble speaking with them as we do with dolphins?

Read full news from source

No comments