According to the artificial intelligence researchers at Facebook’s parent company, Meta has taken a step forward towards achieving its science-fiction dream of automated universal translation.
Sources claimed that machine learning, the AI-supporting technology can translate 204 languages including the Acehnese language of Indonesia and the Chokwe language of Central and Southern Africa with a higher level of translation quality.
These languages have never been simple for computers to translate because of their negligible online presence. Meta and Facebook CEO Mark Zuckerberg praised the AI translation achievement and dubbed it a superpower.
Notably, AI has become the most controversial area in science, especially after the technology recently came under the limelight when a Google engineer was dismissed for claiming that a chatbot could share its thoughts and sentiments.
One of the Meta researchers Professor Philipp Koehn cited that the study presents an excellent effort to expand production-level translation quality to 200 languages. Moreover, a ton of information will be made available so that everyone utilizes this model and retrain it on their own while continuing with their research fields.
Computer scientists who weren't engaged in the project highlighted that although Meta paper claimed to be setting the critical groundwork towards realizing a universal translation system, it was only a minor step on a long and convoluted journey with no visible end in sight.
Measuring machine learning progress is difficult, however, Meta’s AI trial significantly managed to increase translation quality by 44% above the prior state-of-the-art by using a metric called BLEU. Besides translating languages with minimum speakers, the trial is also expected to raise the bar for the quality of the algorithm and improve the underlying language data.
Source Credits:
© 2024 aeresearch.net. All Rights Reserved.