[ExI] AI Translation

[John Clark]

A funny thing happened on November 4, although no public announcement was
made Google Translate suddenly started to do
​ a​
much much better
​ job​
.
​​
Historically machine translation has improved a lot but the rate of
improvement has slowed down over the last 4 or 5 years, until November 4.
And then Boom! For example before November 4 it would translate the
​
Jorge  Borges
​
quote
​"​
*Uno no es lo que es por lo que escribe, sino por lo que ha leído​"​*
​
as:

“
*One is not what is for what he writes, but for what he has read*”

Pretty clunky, but after November 4 it was:

“*You are not what you write, but what you have read*”

People in Japan were among the first to notice
​that something was up and within hours of the change the topic was the # 1
trend on Japanese Twitter. Somebody took a paragraph from a Japanese
translation of
“The Snows of Kilimanjaro”
​ made by a human and asked Google-translate to translate it back into
English. One of the below is pure Hemingway and the other is ​
Hemingway
​ translated to ​
Japanese
​ and then translated by Google back to English, see if you can tell which
is which:​

*"Kilimanjaro is a snow-covered mountain 19,710 feet high, and is said to
be the highest mountain in Africa. Its western summit is called the Masai
“Ngaje Ngai,” the House of God. Close to the western summit there is the
dried and frozen carcass of a leopard. No one has explained what the
leopard was seeking at that altitude."*


*"Kilimanjaro is a mountain of 19,710 feet covered with snow and is said to
be the highest mountain in Africa. The summit of the west is called “Ngaje
Ngai” in Masai, the house of God. Near the top of the west there is a dry
and frozen dead body of leopard. No one has ever explained what leopard
wanted at that altitude".*


The first one is pure
Hemingway
​'s and the second went through 2 translations, one of them make by a
machine. Pretty damn good, the only giveaway is the lack of an article
before the word "leopard". Odd it captured the subtlety but missed the
"the" and a "a".  But compare that with what Google-Translate would have
give you just one day before:

*"Kilimanjaro is 19,710 feet of the mountain covered with snow, and it is
said that the highest mountain in Africa. Top of the west, “Ngaje Ngai” in
the Maasai language, has been referred to as the house of God. The top
close to the west, there is a dry, frozen carcass of a leopard. Whether the
leopard had what the demand at that altitude, there is no that nobody
explained."*

​Granted it remembered to put an article before "leopard" but overall I
think you'll agree it's a greatly inferior translation. So what happened?
It turns out it all started in February, Google decided to switch over from
a symbolic AI system that they and everybody else has been using for
machine translation for the last 30 years to a neural net. They figured it
would take at least 3 years to make the change because they not only had to
write a lot of code they'd also have to make a new chip called a “tensor
processing unit”  to deal with the heavy user volume. But things went much
smother than expected and in mid March they stopped all new work on the old
symbolic AI translation system to concentrate on neural net translation,
and by early November they were finished.

Google thinks a system like they've developed that recognized patterns and
patterns of patterns that worked so well for translation could work equally
well for other things, like going through millions of pages of documents
for legal discovery far faster and cheaper than any human lawyer could. It
should also work to greatly improve image recognition;  and that would put
more than just 3 and a half million truck drivers out of work, neural
networks are much better at finding tumors in X-rays and MRI's than human
radiologist. It seems to me that in the health care field the only two
areas where humans are still better than machines are nursing and surgery.

Translators, doctors, lawyers, truck drivers.... although our civilization
will generate more wealth than ever before people are going to find it
increasingly difficult to find a job.

The New York Times has a good article about all this:

http://www.nytimes.com/2016/12/14/magazine/the-great-ai-awakening.html?rref=collection%2Fsectioncollection%2Fmagazine


John K Clark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20161219/f516952a/attachment.html>