While the term “artificial intelligence” might conjure images of life-like robots walking the streets, the reality is that AI. already exists all around us in much more subtle ways. One such area in which AI is currently undergoing a revolution is in the online translation industry. The work here is bringing what is known as General AI closer every day
While the technology is not yet developed enough to render accurate, contextualized translations on its own, it seems to be emerging as a valuable tool for translations professionals.
Four Types of AI
Presently, the four main types of Artificial Intelligence. include Narrow Artificial Intelligence, Machine Learning, Deep Learning, and General AI.
The oldest and simplest form of AI is known as Narrow AI Narrow Artificial Intelligence refers to technology that can perform simple, rule-based tasks with equal or more accuracy than humans. Most Artificial Intelligence programs currently in use fall under the category of Narrow AI, including Facebook’s facial recognition software.
Next to be developed was Machine Learning, which allows machines to go beyond a set of rules to perform more complex tasks. It involves using algorithms and inputting extensive amounts of data to get machines to make predictions based on the information.
Deep Learning is a much newer form of Artificial Intelligence made to mirror the human brain with a complex network of artificial neural pathways. This model allows the AI to learn from its own errors and use that information to make new gains in areas for which it has not been specifically programmed.
Finally, there’s General AI which has not yet come to fruition but is expected to in the coming years. This type of Artificial Intelligence is what movie-makers and sci-fi writers alike have imagined for years: intelligent, human-like creations that can carry out diverse tasks and learn on their own. Picture C-3PO from Star Wars or Data from Star Trek.
Breakthroughs in Artificial Intelligence and Machine Translation
Advances in the last two decades have allowed researchers to develop Deep Learning technology that creates massive artificial neural networks. These networks can be trained to sift through huge amounts of data, make predictions, and learn from their mistakes, much like a human does when learning a new task.
Progress made with AI has revolutionized machine translation. The rule-based AI previously used by Google Translate was ill-suited for its job, since human languages often have just as many exceptions as they do rules. It’s difficult to program machines to take into account things like sarcasm, humor, or cultural competency.
Recently, though, Google has begun using Google Neural Machine Translation, or GNMT. The neural network uses deep learning to improve its accuracy in the target language. Through this process, a surprising discovery was made: the machine had seemingly created its own language in order to translate between two dialects it had not been explicitly taught.
For example, say the system had been programmed to translate from English to Japanese and English to Korean. The system showed that it could now make the leap to translate ‘directly’ from Japanese to Korean by creating an intermediary language–one that’s not readable by humans. Researchers have dubbed this process “zero-shot translation,” and it has proved to be more accurate and nuanced than rule-based processes.
So, if Google Translate is able to create a perfect translation in under 5 seconds, will human translators be rendered obsolete? Not according to the Bureau of Labor statistics, which predicts that jobs for interpreters will actually grow by 18% until 2026.
Perhaps this is due to the imperfections that still remain in machine generated translations, or perhaps employers are wary of entrusting language, a part of what makes humans human, entirely to machines.
Latest posts by Rae Steinbach (see all)
- 5 Ways to Encourage and Improve an Office Learning Culture - October 26, 2018
- General AI Comes Closer to Reality with Google Translate Discoveries - June 29, 2018
- Artificial Intelligence Might Be the Perfect Party Host - June 6, 2018