Meta Upgrades NLLB AI Model, Cloud Computing Boom, More News

The conversation around the metaverse in the workplace tends to focus on developing the associated technology, particularly communications technology. Yet a fundamental communication problem predates the current phase of metaverse development that vendors are starting to tackle. That problem is language or, more precisely, translating the thousands of languages spoken throughout the world.

San Francisco-based Meta, which has tied its future to the metaverse mast, has just announced an upgrade of its No Language Left Behind (NLLB) AI model which, it claims, will translated 200 different languages far more accurately than any other technology currently on the market. The upgrade has applications well beyond the company’s metaverse ambitions, into its social media platforms and beyond. 

According to Meta, when comparing the quality of translations to previous AI research, NLLB-200 scored an average of 44% higher. For some African and Indian-based languages, NLLB-200’s translations were more than 70% more accurate.

The problem, according to the company, is that AI translation systems of today are not designed to serve the thousands of languages used around the world, or to provide real-time speech-to-speech translation. To truly serve everyone, the machine translation research community will need to overcome three important challenges:

• Translators need to overcome data scarcity by acquiring more training data in more languages
• Overcome the modelling challenges that arise as models
• Find new ways to evaluate and improve on their results.

The AI translation systems of today are not designed to serve the over 7100 languages used around the world, or to provide real-time speech-to-speech translation. Clearly Meta is not the only vendor working on the challenge.

In May, for example, Google announced it added 24 languages to Google Translate with the help of pioneering techniques to reduce noise in the samples of lesser-used languages.

The Google paper, titled “Building Machine Translation Systems for the Next Thousand Languages,” makes clear that the firm will not stop at the 133 languages currently featured on Google Translate.

Both Google and Meta face competition from Microsoft as well as smaller players in the space. All of them have business motivations for their research, not least because the more people use their tools, the better the data they can feed back into the AI loop.

Meta’s European Data Problems

Sticking with Meta, Europe has taken another step toward blocking the transfer of data from Europe to the U.S. this week after Ireland’s data regulator shared a draft order with other EU regulators.

The move follows the issuance of a provisional order by the Data Protection Commission (DPC) in 2020 that blocked the Privacy Shield, the mechanism Meta and other companies use to transfer the data of EU users to the United States. This after Europe’s highest court deemed such data transfer agreements faulty and invalid because of surveillance concerns.

Since then, the European Union and the U.S. arrived at a provisional agreement to permit the data transfer until a permanent arrangement was met. However, the DPC’s probe has continued in parallel.

On July 7, it informed its EU counterparts of a draft of its final decision, although it didn’t go into the specifics of the decision.

The DPC is the EU’s lead regulator of Meta and many other of the world’s largest technology companies, due to the fact that many of the big tech companies have their EU headquarters in Ireland.

Under EU privacy rules introduced in 2018, regulators around the bloc have one month to give their input before a final decision is reached. Any objections, which have regularly been lodged in such cases, could add months to the timeline.

Meta has warned a stoppage will likely leave it unable to offer significant services such as Facebook and Instagram in Europe without a new transatlantic data transfer framework.

The Data Protection Commissioner for Ireland, Helen Dixon told the Irish Times in February that a halt to Meta’s data flows would not immediately hit other big tech companies, but that “there would potentially be ‘hundreds of thousands of entities’ that would have to be looked at.”

This week’s decision will not have an immediate effect on Meta’s services. The draft decision triggers the process under Article 60 of the GDPR. Article 60 provides the other Data Privacy Agencies (DPAs) in Europe four weeks to raise issues with the decision, something that will likely occur before a definitive ruling is made later this summer.

Earlier this year, Meta outlined what would happen if the EU and U.S. failed to agree upon a new privacy agreement in an SEC filing: “If a new transatlantic data transfer framework is not adopted and we are unable to continue to rely on SCCs or rely upon other alternative means of data transfers from Europe to the United States, we will likely be unable to offer a number of our most significant products and services, including Facebook and Instagram, in Europe.” 

The long-term impact of this decision is impossible to tell at this point, but it holds the potential to cut off international organizations from a great deal of the data they need to operate across geographies.

Cloud Computing Is Booming

Cloud computing exploded during the pandemic, which should surprise no one given the number of people who were working from home who needed easy access to tools and data. Data out of Needham, Mass.-based IDC established just how big an impact the pandemic had on the space.

According to the IDC’s Worldwide Semiannual Public Cloud Services Tracker, spending on all kinds of cloud computing environments — including compute, storage infrastructure as well as dedicated shared environments — have increased by 17.2% in the first quarter to $18.3 billion.