Machine Translation

Is there a future for translation by computer? Can we hope that in the near future computers will take the place of translators?

LeTonBeauDeMarot is a must-read book if you're interested in language translation (by humans, or especially by artificial intelligence). There's also a few nuggets about the trickiness of self-reference (see QuineProgram). Hofstadter does say that a very few specific pieces of text "are probably forever untranslatable"[p.444], including this book itself [p.450]. -- DavidCary

I believe that isotopic context analysis of a paragraph can provide enough information to deal with polysemic words. And a system of simple sentence analyses already can decide about the uses of expressions and idioms. When I was 12, I was already experimenting with computer-translation on Latin texts, just as well I was programming the verbal conjugations of French verbes. (Creating in the process of doing so all the accentuated letters I needed on my CommodoreSixtyFour. And also for the printer - I created my own word processor after reading a book on the different function a good word processor had - because that wasn't standard neither at that time. -- PieterJansegers

It all depends on solving the "common sense" problem. Computers lack common sense. Humans spend years learning millions of seemingly trivial facts that can be drawn upon during communication. The CycProject was one attempt to solve this problem. Cog is another. Eventually someone will figure it out because humans aren't magical, they're just complicated. It's only a matter of time before we can build an emulation of one well enough that it can understand what we say. It may take another 10,000 years. 50 years is nothing compared to how long it took humans to evolve. Look at how long it took us to build flying machines.

There's been life on this planet for over 3 billion years. I doubt anything we do will stop that. No one knows how much oil is in the ground. No one knows when it will run out. People will survive without oil, one way or another. People have been saying there's nothing left to invent for at least 100 years, and yet new things are still invented. There's plenty left to be invented. Our brains and our bodies are machines. There's nothing magical about them. Computers are the machines that can imitate any machine. I feed ideas to them all the time. They aren't clever enough or experienced enough to understand them in natural language yet, but there's no reason I can see that won't change given enough time.

Planes didn't exist 100 years ago (this was written before Dec 17th, 2003). Liquid fueled rockets didn't exist. Thermonuclear bombs didn't exist. Transistors didn't exist. DNA was unknown, and there were no gene sequencers. When I started programming (a little over 20 years ago) I was happy to get 16k of RAM and a 1MHZ processor for $500. Today I can get 32,000 times as much RAM and a processor running 15,000 times faster for the same money. I can edit digital video on my $500 PC today. There were no digital video editors 20 years ago. None of us can predict what computers will be like in 100 years. -- EricHodges


See also: BabelFish


CategoryArtificialIntelligence


EditText of this page (last edited October 25, 2005) or FindPage with title or text search