MARK Zuckerberg and his super-company Meta have revealed an artificial intelligence model that can translate more than 200 languages.
The AI program is part of Meta’s No Language Left Behind, a machine translation workshop with applications on the internet and in the metaverse.
2
2
Meta’s model is called NLLB-200 and it is versed in languages that are used internationally or by a select few.
The company published a 190-page paper on NLLB-200 and its motivations, which notes that languages fading from usage represent a global cultural decline.
“Many low-resource language speakers remind us that machine
translation could be a critical tool in promoting language and cultural preservation,” the paper states.
While Meta touts the model’s potential for inclusivity, there is a prominent fear that AI will eventually become too intelligent to contain.
But Meta’s NLLB-200 program is not a decision-making AI program that can result in the immediate loss of life like an autonomous weapons system.
The paper notes that NLLB-200 is naturally of use to Westerners and common language speakers traveling but also represents a democratization of the internet.
“A Tigrinya speaker notes that in Ethiopia, where less than 20 percent of the country has internet access, actual access to what the web offers is even more restricted due to the lack of quality translation,” the paper writes.
But the AI program is not without its risks – the translated internet could open up communities to more advanced censorship, scams, or hate speech.
Most read in Tech
NLLB-200 is powered by thousands of hours of interviews, videos, and conversations with real human speakers and translators.
The datasets underwriting the model are publicly available through Meta.
NLLB-200 is geared to “train” in translation on these millions of examples and is then checked against human-translated phrases for quality assurance.
The AI model landscape is a competitive industry with lots of money on the table.
Meta, Google, and other major tech players are in a race to develop AI models and situate them for use on their platforms.
This post first appeared on Thesun.co.uk
