Fascinating. They did not release the new program, saying it was too dangerous.
https://medium.com/@ageitgey/deepfak...s-5e057ebd697d
Has anyone ever tried to build a machine translation app or software program?
I did fool around with some algorithms, but the first problem you run into is the part of speech.
The word "read" can be a noun ("this article was a good read") or a verb ("I read the book") but it could be present past or future tense - so translating it into Chinese is no problem (it would be the same character in all tenses) but in most other languages, you need to know the tense, which can probably be ascertained by examining the words before or after the word "read" ("have read" or "will read"), but "I read the book" does not indicate if you are reading the book now or have already read the book. I suppose this is less of a problem in Russian because, quite often, the word ending would indicate the part of speech, the tense and other information about the word.
Then there is case. "I walk to the store." If you can figure out "store" is a noun and not a verb, the next useful bit of information is that it is not the subject of the sentence, but an object. And in Russian there are 6 cases, one more than in English, which presents another problem. Once you have the case, sentence structure and part of speech information, then you can "reverse engineer" the sentence into your target language. That's about as far as I got. The best translation software I ever saw was "Translate it!" and it was written in DOS, many moons ago. But it seems machine translation is now going in other directions.
I think Google and Yandex translates by using "brute force" algorithms - comparing the sentence or phrase you enter to millions of stored, already-translated sentences and phrases. Also, Google likes to "predict" what you are looking for (this is one area where Google interfered in our society, by altering the user's "search predictions". If you search for "Republican Senator" Google's first "suggestion" is "Republican senators in trouble", and in a search for "Democrat Senators", Google suggests "Democratic Senators 2020"). And of course, Google's search results themselves are man-and-machine manipulated - if you search for "election fraud" you will get one article explaining where and how the fraud occurred, and a thousand articles proclaiming there was no fraud.
But "prediction " would play a small role in a good machine translation system, it would be good to save you time in writing,. like on your smart phone that predicts the next word in your text message.
BTW, where did Russians get "SMS text message" from? Why did they not just say "text message"?