SoylentNews is people
SoylentNews
from the google-is-so-smart-it-can-maek-typos-for-us dept.
The NY Times covers in a very in-depth article Google's contribution to the recent revolution in deep learning through its application in Google Translation. A great read that covers the journey from new theories to practice in less than 10 years. This piece is surprising very technical as it tries to explains the history, the people and the technology behind the recent AI revolution. Take the time to read the full story here.
It is, in fact, three overlapping stories that converge in Google Translate's successful metamorphosis to A.I. — a technical story, an institutional story and a story about the evolution of ideas. The technical story is about one team on one product at one company, and the process by which they refined, tested and introduced a brand-new version of an old product in only about a quarter of the time anyone, themselves included, might reasonably have expected. The institutional story is about the employees of a small but influential artificial-intelligence group within that company, and the process by which their intuitive faith in some old, unproven and broadly unpalatable notions about computing upended every other company within a large radius. The story of ideas is about the cognitive scientists, psychologists and wayward engineers who long toiled in obscurity, and the process by which their ostensibly irrational convictions ultimately inspired a paradigm shift in our understanding not only of technology but also, in theory, of consciousness itself.
(Score: 2) by SomeGuy on Sunday December 18 2016, @01:23AM
Summary of all AI in three words: "KILL ALL HUMANS"
(Score: 0, Troll) by Ethanol-fueled on Sunday December 18 2016, @03:09AM
Summary of Google AI in three words: "MUST ELECT HILLARY"
(Score: 0) by Anonymous Coward on Sunday December 18 2016, @04:00AM
As suggested in tfs, I read the full article just now. It's a good yarn and I second the recommendation to read it all. The conclusion is that machine learning (aka, neural networks running on huge hardware) is coming to a profession near you and near me. It still takes tweaking to get things to work, but less with experience.
(Score: 2) by Gaaark on Sunday December 18 2016, @01:41AM
Soylent's AI revolution from from from from theory to practice in 2 2 2 years years.
WWWTFFF? ? ;)
--- Please remind me if I haven't been civil to you: I'm channeling MDC. I have always been here. ---Gaaark 2.0 --
(Score: 0) by Anonymous Coward on Sunday December 18 2016, @02:09AM
A Japanese friend visited and we went to a couple of museums. He wrote up our little road trip (in Japanese) for a special interest magazine in Japan, and sent me an electronic version. I ran through Google Translate a couple of years ago and got mostly gibberish, with just enough recognizable words so show that it was trying. Just now I repeated the exercise after reading that Translate has been improved. Well, this time I also got gibberish, with a few more recognizable words and maybe a few sensible phrases of a few words in length. This is from two pages of Japanese characters.
This new AI version might be good for Hemingway (from the link), but not for my friend who is a well known writer in his field.
(Score: -1, Troll) by Anonymous Coward on Sunday December 18 2016, @02:31AM
Maybe Japanese is a shit language.
(Score: -1, Offtopic) by Anonymous Coward on Sunday December 18 2016, @03:13AM
Maybe you are a shitposter.
(Score: 1) by anubi on Sunday December 18 2016, @07:14AM
I think this illustrates the complexities and nuances of translation.
Our best minds today programming the most advanced computer systems have demonstrated thus.
Now, consider the efforts of translating our Holy Books from their Aramaic, Hebrew, and Greek roots.
Considering this, I sure wish we would stop fighting over religion.
But then, I also believe that we are really fighting over economic control over who pays tribute to who, and to save face, religion is used as the scapegoat.
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 0) by Anonymous Coward on Sunday December 18 2016, @10:46PM
"our"
"we"
(Score: 2, Informative) by Anonymous Coward on Sunday December 18 2016, @03:19AM
That's a misconception about language - especially when it comes to poetic writing, but to some degree with all. When it is translated, inevitably, at best, a subset of possible interpretations is captured by the new language. Translating that subset back to the original language means you are getting a subset of all possible interpretations of something in an entirely different language. This new subset is intersected with the original intended meaning, but generally very different and incompatible.
(Score: 2) by HiThere on Sunday December 18 2016, @08:07PM
For really short phrases out of context, that's true. Otherwise...not. You don't get the actual connotations included in the original message, but you get the same basic meanings. This isn't invariably true, of course. I can't imagine a translation of Finnegan's Wake (James Joyce) even into Dutch, German, or Irish, much less into any other language. In fact there are parts that need translation into US American, and other parts that need translation FROM US American into British.
But the fact that edge cases exist doesn't mean that the basic process is invalid. To give an old classic example, the translation of "out of sight, out of mind" into "invisible idiot" or "distant mad man" is quite reasonable if you don't have context or know the idioms. But those things are part of a good modern translation program. One of the problems is that there are so many idioms. Another, of course, is that frequently context requires a good model of physics and another good model of emotional interactions. And "good" isn't a well-defined term in either of those cases.
Javascript is what you use to allow unknown third parties to run software you have no idea about on your computer.
(Score: 2) by darkfeline on Monday December 19 2016, @08:28AM
Can you post it? My experience with Chinese and Japanese is that Google Translate has gotten a lot better. It still struggles with more academic style writing, but at least for more conversational stuff it's very good (which is surprising given how horrible Translate used to be with Chinese).
Consider the following from Wikipedia: 前項のように、筆は使うほどに本来の筆の持つ能力が引き出されてくるが、それには墨の選び方や洗い方も大事になってくる。
Translate says: As mentioned in the preceding paragraph, the ability of the original brush to draw out as much as you can use the brush will be drawn out, but how to choose ink and how to wash will become more important.
It's still a bit wonky, but I wouldn't call that gibberish.
(The actual translation: As described in the previous paragraph, while the pen's potential is drawn out as it is used, the selection of ink and how the pen is washed is also important.)
Join the SDF Public Access UNIX System today!
(Score: 0) by Anonymous Coward on Monday December 19 2016, @02:44AM
As one of the elderly folks on these here interwebs, I'm constantly entertained by how old news becomes new again.
Back in the 1970s, IBM produced a translate computer (yes, a refrigerator sized dedicated machine) to convert English to Chinese and vice versa. When the CEO arrived to test it. he entered "Out of sight, out of mind." The machine translated into Chinese, which he couldn't read, so he entered that into the machine to translate back into English. The machine responded "Invisible idiot."
He instructed them to keep working on it. And here we are.
To be fair, my CS professors also told us that voice recognition would not, could not happen in our lifetimes. They laughed out loud at Lorne Green talking to his terminal on Battle Star Galactica.
Hello, Siri?
(Score: 2) by darkfeline on Monday December 19 2016, @08:33AM
Nice anecdote, but we've kept working on it.
Google Translate: Out of sight, out of mind -> 看不见，不在乎。-> Can not see, do not care.
That's pretty good, in my opinion. If anyone has some Chinese translations before the new Translate AI, it'd be fun to look at, as it used to be completely unusable.
Join the SDF Public Access UNIX System today!
(Score: 2) by takyon on Wednesday December 21 2016, @04:48AM
I think there is some incredibly important and new news here. Google Translate made more improvements by switching to their new machine learning and TPU approach (TPUs being a story in and of themselves [soylentnews.org]) in just a few months than in the preceding decade (they gave a metric for measuring the improvement - the BLEU score). And those improvements are coming to a wide range of similar computer-human-world interaction tasks. The massive improvements being made to Google Translate are being made despite the fact that no intelligence is involved. It's still just a pattern-finding algorithm.
The TPU and machine learning revolution over at Google might be the biggest tech story of the decade, and it's unfolding right in front of our eyes.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]