Machine translation is based on "playful theft"
Thread poster: Abba Storgen (X)
Abba Storgen (X)
Abba Storgen (X)
United States
Local time: 12:21
Greek to English
+ ...
Feb 28, 2021

In Jaron Lanier's interview on YouTube with full video title "The REAL Social Dilemma w/ Jaron Lanier | Full Interview | The Jordan Harbinger Show Ep 156", he mentions at around time point 31.00 and on, that these AI machines need data continuously from people to feed the machine - so far we have known that. In other videos with general titles "who owns the future" and "the promise of AI" etc, he goes a little deeper on the matter of translations, explaining that the machines collect data contin... See more
In Jaron Lanier's interview on YouTube with full video title "The REAL Social Dilemma w/ Jaron Lanier | Full Interview | The Jordan Harbinger Show Ep 156", he mentions at around time point 31.00 and on, that these AI machines need data continuously from people to feed the machine - so far we have known that. In other videos with general titles "who owns the future" and "the promise of AI" etc, he goes a little deeper on the matter of translations, explaining that the machines collect data continuously from translators, in the background (obviously he refers to the Google API translators use all the time), feed them into a "general corpus" and then process them etc.
Which of course also explains why the Google machine API produces vastly different results when asked to translate a sentence in all capital letters in various languages - the all caps sentences are not good at all, because there are not enough sentences in all caps "in the corpus" provided by translators yet to form a sufficient "translation memory". This proves that MT is not smart at all, it's just a gigantic translation memory with a few tweaks here and there, but it is almost entirely dependent on continuous new input from translators, which is steals through the API, while claiming it's not doing it.
Google's claim in the API that the application does not collect source data is invalid (how can it provide the translation without collecting the data, and why do I get sometimes random company names in certain sentences that do not exist in the source?). Google's claim is as good as their claim that their home security system was not spying on its customers (it was found that it contained a secret microphone, not existent in the schematics or the manual, which then Google promised to deactivate... see the documentary "Surveillance Capitalism" on this and other topics).
-----
Of course some colleagues will say "well that's tech, deal with it", but I do not see AI replacing teachers (they have strong unions and political clout) or the vast army of public employees, with 50% of them easily replaceable by much older tech than AI. My laptop could easily replace 15 bureaucrats and 50 librarians in the public sector in my area, but I'm supposed to care about their jobs, while nobody is supposed to care about mine.
-----
In essence, what I say, is that while other sectors of the economy have fiercely protected their members, declaring that they will protect their interests at all costs (even at the cost of others), and raising revolutionary banners every time someone proposes to delay their salary increases by 1 year, translators are among those who are 100% unprotected and subject to whatever technological change.
Collapse


Francesco Volpe
Robert Rietvelt
Arjan van den Berg
Oriol Vives (X)
Robert Edison
 
CARL HARRIS
CARL HARRIS
United States
Local time: 12:21
Member (2013)
English to French
+ ...
Sophisticated algorithms Feb 28, 2021

The term Artificial Intelligence (AI) is used loosely however, the reality is that AI has not yet been invented. It's still a compilation of sophisticated algorithms.

Kay-Viktor Stegemann
Robert Rietvelt
Oriol Vives (X)
Matthias Brombach
Kirill Rodionov
ahartje
Laura Kingdon
 
Anton Konashenok
Anton Konashenok  Identity Verified
Czech Republic
Local time: 18:21
French to English
+ ...
The outcome is bound to be poor Feb 28, 2021

Today's AI translation programs suffer from two inherent problems:
- They need to be trained, and their output is very inlikely to be any better than the training input. Excellent translators are few and expensive, so all but the most high-budget systems are trained on mediocre data. Thus, they are condemned to eternal mediocrity.
- Artificial intelligence tends to suffer from "artificlal arrogance", or "artificial sciolism": it just doesn't say "Sorry, I don't know".


Peter Shortall
Arabic & More
Anne Spitzmueller
Oriol Vives (X)
John Fossey
Laura Kingdon
Irene (Renata) Liapis
 
Arjan van den Berg
Arjan van den Berg  Identity Verified
Local time: 19:21
Member (2010)
English to Dutch
+ ...
"This proves that MT is not smart at all, it's just a gigantic translation memory with a few tweaks" Mar 1, 2021

Another proof for the statement "This proves that MT is not smart at all, it's just a gigantic translation memory with a few tweaks":
Run a complicated T&C with a lot of legalize through Google translate and then do the same with a simple story with some dialogue you wrote yourself. The translation of the T&C will be better. You will also see that the same terms will be translated differently throughout the legal document. This shows that different TM's are used for different parts of the
... See more
Another proof for the statement "This proves that MT is not smart at all, it's just a gigantic translation memory with a few tweaks":
Run a complicated T&C with a lot of legalize through Google translate and then do the same with a simple story with some dialogue you wrote yourself. The translation of the T&C will be better. You will also see that the same terms will be translated differently throughout the legal document. This shows that different TM's are used for different parts of the document. In Dutch for example 'business' can be translated in (at least) two ways. You will see that both versions will be found in the 'translated' document.
Collapse


 
Adieu
Adieu  Identity Verified
Ukrainian to English
+ ...
You meant legalese Mar 28, 2021

"Legalize" is the rallying cry of proponents of importing your country's light drugs policy.

Everything else you say is spot-on, though, since MT excels at standardized and highly formulaic bureaucratic or legal language.

Then again, it might just fail spectacularly if you copy a typical contract and sparingly throw in a few strategically placed meaning-reversing substitutions or outright negatives (just not too many, keep it at 99+% match and see what happens)

[
... See more
"Legalize" is the rallying cry of proponents of importing your country's light drugs policy.

Everything else you say is spot-on, though, since MT excels at standardized and highly formulaic bureaucratic or legal language.

Then again, it might just fail spectacularly if you copy a typical contract and sparingly throw in a few strategically placed meaning-reversing substitutions or outright negatives (just not too many, keep it at 99+% match and see what happens)

[Edited at 2021-03-28 15:58 GMT]
Collapse


 
Sadek_A
Sadek_A  Identity Verified
Local time: 21:21
English to Arabic
+ ...
..... Mar 28, 2021

Sorry, but this original post, when examined side by side with your clear stance in the PROAct thread, only succeed in perplexing.

Are you with, or against, protecting the interests of translators?

A government, the US one in this case, is making a fist -through that Act- to punch away what it can of the current abuse and exploitation. Still, there you said government should NOT interfere.

If governments are not allowed to interfere, and protection laws are
... See more
Sorry, but this original post, when examined side by side with your clear stance in the PROAct thread, only succeed in perplexing.

Are you with, or against, protecting the interests of translators?

A government, the US one in this case, is making a fist -through that Act- to punch away what it can of the current abuse and exploitation. Still, there you said government should NOT interfere.

If governments are not allowed to interfere, and protection laws are not allowed to be enacted, then WHO will protect, and HOW?
Collapse


 


There is no moderator assigned specifically to this forum.
To report site rules violations or get help, please contact site staff »


Machine translation is based on "playful theft"







Wordfast Pro
Translation Memory Software for Any Platform

Exclusive discount for ProZ.com users! Save over 13% when purchasing Wordfast Pro through ProZ.com. Wordfast is the world's #1 provider of platform-independent Translation Memory software. Consistently ranked the most user-friendly and highest value

Buy now! »
CafeTran Espresso
You've never met a CAT tool this clever!

Translate faster & easier, using a sophisticated CAT tool built by a translator / developer. Accept jobs from clients who use Trados, MemoQ, Wordfast & major CAT tools. Download and start using CafeTran Espresso -- for free

Buy now! »