[ad_1]
Meta CEO Mark Zuckerberg said on Wednesday that his organization can now translate 200 languages into each other, doubling the number in two years. – Photo Factory COFFRINI / AFP
Sunday, July 10, 2022 9 ፡ total 14 AM MYT
July 10 – Seppi, a Peruvian woman who only knows Ketchup, can talk to a man in South Africa.
The universal interpreter is the core of science fiction, but Google, Meta, and others are locked in a battle to work with as many languages as possible with their AI models.
Meta CEO Mark Zuckerberg said on Wednesday that his organization can now translate 200 languages into each other, doubling the number in two years.
A.D. Meta creation in 2020 was to cut ties with English – a long-running language due to the availability of many sources.
Instead, meta models go straight from Chinese to French without going into English.
In May, Google announced a major step forward in adding 24 languages to Google Translate after pioneering techniques to reduce noise in samples of smaller languages.
Sepedi and Quechua were among them – so Peru and South Africans can now meet but only in writing.
Researchers warn that the dream of a real-time conversation interpreter is still a certain way.
Quantity and quality
Both Google and Meta are motivated by their research, at least when more people use their devices, the better the information will be returned to the AI loop.
They are also competing with a popular web-based tool that focuses on fewer languages than rivals Microsoft and DeepL.
Automated translation test is “especially useful” for Facebook because it contains hate speech and inappropriate content that needs to be investigated, researcher Fran ኢois Ivan told AFP.
The tool helps to identify the content of English-speaking moderators, for example, in many other languages.
Meta promotion videos, however, focus on the release features of the technology – amateur chefs appear at their fingertips from far and wide.
But both companies are at the forefront of AI research, and both have accompanied their ads with academic papers that highlight their aspirations.
Google has expressed dissatisfaction with the 133 languages published by Google Translator under the heading “Construction Machine Translation Systems for the Next Thousand Languages”.
However, like Clichy, quantity does not always mean quality.
European priority
“We shouldn’t expect 200×200 language pairs to be of the same quality,” said Ivan, a Facebook model partner.
European languages, for example, are always useful because they have more reliable sources.
According to regular users such as Google Translator and other automated programs, the built-in text may be a robot and there are few errors.
While this may not be a problem for fake restaurant menus for daily use, it does limit the use of those devices.
“You can’t make a mistake when you translate a assembly kit for a fighter jet,” said Vincent Godard, head of French technology company Sistra.
And the ultimate goal of cracking is to create a tool that can easily translate the spoken word.
“We haven’t arrived yet, but we are working on it,” said Antoine Bordes, who runs the Meta AI Research Laboratory Fair.
He said the Meta speech translation project currently works in very few languages.
“But the interest will be connecting the two projects, so one day we will be able to speak in 200 languages with words, emotions, accents,” he said. – ETX Studio
[ad_2]
Source link
