Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Global Crossing - GX (formerly GBLX) -- Ignore unavailable to you. Want to Upgrade?

To: Bill Fischofer who wrote (4522)	2/24/2000 9:25:00 PM
From: J Gunn	Respond to of 15615

Actually, the future requirement, will be "Do you exist, and think." Speaking and seeing and hearing disabilities have already been overcome by technology.

To: Bill Fischofer who wrote (4522)	2/24/2000 9:57:00 PM
From: Wyätt Gwyön	Respond to of 15615

OT, Bill, I remember several years ago, like 1997, Gordon Moore predicted in Fortune Magazine that processors would be so powerful by the year 2000 that you would be able to just click a button and have Fortune translated into 50 different languages. Actually, people were predicting the automation of language translation back in the early 90s, and the early 80s, and the early 70s, and so on. Automated translation is known as "machine translation", or MT. MT does in fact exist today. In addition to low-level consumer packages (which tend to be confined to translation between European languages for personal correspondence and so on), there are industrial-type programs as well. But their uses are pretty narrow in focus. For example, a co. like Kodak or 3M might want to check over a bunch of Japanese patents (Japanese issue huge numbers of patents, and as you might guess, it's a bit harder to automate the translation of this language into English than something like French or German). So they could use MT to translate reams of patents, or just the claims. They could then sift through the gobbledygook and decide if any of the patents look like they have something of interest. In that case, they would then pay a human translator to actually give them a finished product. Between pure MT and pure human translation is a nebulous land of "translation assistant programs", most of which seem a waste of time to me, but some like them. A lot depends on the type of text you are working with and what level of end product you will be satisfied with. In any case, while there are currently legitimate uses of MT in commercial settings, I think it'll be a long time before human translators are out of the picture, especially when working between two languages that have radically different syntaxes and lexicons. As I see it, the problem with the idea of a "total MT solution" (i.e., a killer program that could provide you with on-the-fly reliable translation in any kind of setting) is that languages are inherently ambiguous. To give you a simple example Chomsky brought up ages ago, think about the ambiguity in the following sentence: "Flying planes can be dangerous." Because this sentence can mean different things, you have to have a context greater than the sentence itself in order to derive the meaning intended by the speaker (or writer). This isn't a big problem in normal conversation in one language--we unconsciously intuit the probable meaning. But, if you want to translate that sentence into another language which does not allow you to preserve that ambiguity (as in the case of Japanese or Korean), then you're going to have to make a judgment call as to what you think that sentence means. You have to choose door #1 or door #2. The more you look at languages, the more you see how pervasive this ambiguity is. Translators therefore have to make these judgment calls all the time. But there is no simple algorithm that you can program into a computer to make such judgments, because it is "outside the box", i.e., it is not even an input to the problem. The most you could expect from an algorithm is that it would recognize its inability to come up with the correct answer. A real killer MT program would, in my mind, be a conscious machine--like "Data" in Star Trek. I believe you need consciousness in order to understand the total range of contexts that people refer to in their communications--cuz this is the range (i.e., everything), that impacts how we use language. I don't think a Pentium III is gonna cut it. JMHO.