“First the machines will do a lot of jobs for us and not be super intelligent. That should be positive, if we manage it well. A few decades after that though, the intelligence is strong enough to be a concern.”
The year is 2037, at the UN headquarters in New York, diplomats and delegates from all over the world get ready for a session, put on their head phones, turn on their translating devices, and the session begins. A small difference to our current state – no human is interpreting the words that are spoken. The translation device has been designed to automatically convert words from one language to another. And the translation is immaculate, accurate, without even the slightest chance of human error, forget inaccuracies. It this vision completely unfathomable?
Will machine translation take over human translation in this century? Should translators be concerned?
Presently, machine translation is not perfect.
Machine translation, also abbreviated as MT, usually does not create an accurate translation of the text because recognition of entire phrases in a target language is relatively still at an initial stage. The quality of engines has reached the point where using them leads to proven productivity gains. Currently, machine translation makes our work easier rather than threaten to take over our jobs. In fact, they enable easy post-editing and deliveries of translations at a quick turnaround time. This is particularly true for translations such as contracts, software documents, e-learning content, etc. According to translationjournal.net, as machine translation is being constantly improved, many, if not most human translators will eventually become just post-editors of MT.
However, I came across this quote by Ray Kurzweil, who is the author of The Age of Spiritual Machines:
“As long as there is an Artificial Intelligence (AI) shortcoming in any such area of endeavour, sceptics will point to that area as an inherent bastion of permanent human superiority over the capabilities of our own creations.”
Is it right to believe that we hold a human superiority over machines? Yes, Kurzweil tells us, as long as there is an AI shortcoming. Artificial Intelligence is precisely what brings me to the subject of Deep Learning.
Deep Learning is a new area of Machine Learning research. Its objective is to move Machine Learning closer to one of its original goals: Artificial Intelligence. With massive amounts of computational power, machines can now recognize objects and translate speech in real time. Artificial intelligence is finally getting smart.
“Within several decades, information-based technologies will encompass all human knowledge and proficiency, ultimately including the pattern-recognition powers, problem-solving skills, and emotional and moral intelligence of the human brain itself.”, adds Kurzweil.
According to a study by Kevin Duh, from Nara Institute of Science and Technology in Japan, the flowchart for a standard process in machine learning is as follows:
raw data -> hand-engineered features -> trainable classifier -> decision
With Deep Learning, the flowchart has been modified in the following manner:
raw data -> TRAINABLE FEATURES -> trainable classifier -> decision.
An efficient training algorithm is now proposed and deep architectures in machine learning and artificial intelligence are becoming more and more real.
This whole new training procedure consists of 4 phases:
1) An autoencoder for the source words is learnt
2) An autoencoder for the target words is learnt
3) The autoencoders are connected by a top connecting layer
4) Fine-tuning the weights through the entire network
(An autoencoder is an artificial neural network used for learning efficient coding.)
So what does this really mean? What is the forecast for the 21st century? Once again, the words of Kurzweil explain it best:
“It is reasonable to expect the hardware that can emulate human-brain functionality to be available for approximately one thousand dollars by around 2020. (…) The software that will replicate that functionality will take about a decade longer. However, the exponential growth of the price-performance, capacity, and speed of our hardware technology will continue during that period, so by 2030 it will take a village of human brains (around one thousand) to match a thousand dollars’ worth of computing. By 2050, one thousand dollars of computing will exceed the processing power of all human brains on Earth.”
Returning to our initial question: should we be concerned?
Bill Gates recently expressed that he believes we should be worried about Artificial Intelligence. Stephen Hawking believes the same. And so do I. I believe this vision may soon become a reality. I believe “spiritual machines” are a real threat. But the debate is still on. With regard to the translation field, some believe machine translation is never going to replace human translators. For these simple reasons, says Nataly Kelly, on huffpost.com:
a) It’s tough to get good translation, even from perfectly bilingual human beings
b) Translation quality is highly subjective
c) There are too many languages out there
And finally, this is what I think: this is the age of globalization and digitalization. There is a lot of work out there for all of us. A few weeks ago, I was at the Expolangues fair in Paris. This is the third year I attend these conferences. It is encouraging to see that translators are continuously needed in different fields of translations, ranging from the Court of Justice to new technologies such as Surtilting. The trick is to reinvent ourselves. We must learn to adapt. We must find new ways to work alongside new technologies. And we can turn our translation business into a workable profitable solution.
The characteristics of humans, since the beginning of times, have been one of survival and the ability to adapt. So if we are run by highly-developed spiritual machines in a few decades, does that mean we would become highly-evolved spiritual beings, thus rising above mere machines?