Automatic universal translation has been a science fiction fantasy for quite some time. According to a new paper by artificial intelligence researchers at Facebook’s parent company Meta, this objective has been advanced.
It’s the latest news from the controversial field of artificial intelligence, which could potentially revolutionize translation techniques.
The paper demonstrates that machine learning, the technology underlying artificial intelligence, can translate 204 different languages, twice as many as has ever been attempted, with a higher level of accuracy than ever before.
This includes more than a hundred languages that are rarely spoken, such as the languages of the Acehnese people of Indonesia and the Chokwe people of Central and Southern Africa, which have always been difficult for computers to translate due to their limited online presence.
Mark Zuckerberg, CEO of Facebook and Meta, hailed the achievement, calling AI translation a “superpower,” and the researchers were only slightly less ecstatic.
It’s the most recent advancement in artificial intelligence, a controversial field of science that has recently been in the spotlight after a Google engineer was placed on leave for claiming a chatbot could express emotions.
Professor Philipp Koehn of Johns Hopkins University, one of the 38 academics and Meta researchers who collaborated on the work, said, “The paper presents impressive work to push production-level translation quality to 200 languages”
Many resources will be made available so that anyone can use this model and retrain it on their own, fostering research in this area.
So, yes, I believe it is a major issue.
Despite the paper’s claim that it was “laying the important groundwork for the realization of a universal translation system,” computer scientists who were not involved in the project emphasized that it was merely a small step on a long, winding road with no clear destination.
An impressive feat of engineering
Dr. Alexandra Birch-Mayne, Lecturer in Natural Language Processing at the University of Edinburgh, stated that the paper’s central machine learning technique, a model known by the baroque term Sparsely Gated Mixture of Experts, was not new.
According to her, its most significant contribution consisted of collecting, cleaning, and presenting new data on languages that did not appear widely on the internet, the primary source of data for machine translation.
“It is a remarkable feat of engineering. It is not necessarily a scientific breakthrough in terms of fundamentals “.
In addition to translating languages with fewer speakers, the paper claimed to have established a new standard for translation quality.
Mark Zuckerberg is personally being sued over the massive Cambridge Analytica data breach scandal.
Data and algorithms will be made accessible to the public.
Measuring progress in machine learning is a difficult task, but using a metric known as BLEU, the Meta paper improved translation quality by 44% over the prior state of the art.
Dr. Diptesh Kanojia, Lecturer in Artificial Intelligence for Natural Language Processing at the University of Surrey, stated, “BLEU is an imperfect metric.” BLEU scores are, however, a standard practice in natural language processing research.
“From a purely statistical standpoint, this improvement of 44 percent is quite significant.”
Although the work will be used to improve Facebook’s software, the language data and the algorithms used to translate it will be made available to the public, giving other researchers access to authoritative datasets on languages such as Eastern Yiddish, Northern Kurdish, and Cape Verdean Creole for the first time.
Importantly, the Meta researchers found native speakers to review their translations, a time-consuming task that safeguards both the algorithm’s quality and the language data it uses.
“Contributing to one’s community is an admirable trait. They are not necessarily the originators of this trend, but they are adhering to best practices “Dr. Birch-Mayne noted the limitations of the effort, which involved native speakers from Europe and the United States rather than the countries of origin of the languages.
Some researchers criticized the paper’s release without peer review and accused Meta of engaging in “peer review by media.”
Professor Koehn defended the method, stating that it was “common practice in the field… for better or worse” and that it aided in accelerating the dissemination of research findings.
Amazon’s Alexa will soon be able to read your stories in the voice of a deceased family member.
Recent progress in machine learning
The paper is one of several recent advancements in machine learning, which is advancing at a much faster rate than researchers anticipated. Google released a model last week that solved a third of MIT undergraduate math problems with 50 percent accuracy, a significant performance boost.
However, even though each breakthrough inspires speculation about new forms of consciousness, the majority of experts in the field believe that AI systems are neither sentient nor intelligent, claiming that they merely imitate the data they are given. A robot rebellion is not likely.
The greater danger of AI systems is that they will lead to disaster by giving humans false confidence in their still very limited abilities – a very real possibility given the sensitivity of the tasks that could potentially involve translation at Facebook, which has been criticized in the past for failing to employ native-speaking moderators to identify calls for violence on its platform.
Mr. Zuckerberg promised that “the advancements here will enable more than 25 billion translations per day across our apps,” which Facebook said could be used for detecting harmful content, securing elections, and preventing online sexual exploitation.
