Samsung to Add Real-Time Translation to Smartphone Model

Samsung Galaxy Z Fold5 (L) and Z Flip5 (R) smartphones are displayed at a Samsung Electronics store in Seoul on October 31, 2023. (AFP)
Samsung Galaxy Z Fold5 (L) and Z Flip5 (R) smartphones are displayed at a Samsung Electronics store in Seoul on October 31, 2023. (AFP)
TT

Samsung to Add Real-Time Translation to Smartphone Model

Samsung Galaxy Z Fold5 (L) and Z Flip5 (R) smartphones are displayed at a Samsung Electronics store in Seoul on October 31, 2023. (AFP)
Samsung Galaxy Z Fold5 (L) and Z Flip5 (R) smartphones are displayed at a Samsung Electronics store in Seoul on October 31, 2023. (AFP)

Samsung Electronics will roll out a real-time call translation service using AI technology next year, the firm told AFP Friday, claiming it would make conversing across languages as easy as "turning on closed captions".

Samsung is among a host of tech companies spending billions in a global artificial intelligence gold rush, which has so far seen chatbots that mimic human speech and software that can generate lifelike images and videos.

The firm, the world's largest smartphone maker, says the new real-time translation feature will be incorporated into its new Galaxy flagship model, which will launch early next year.

It will enable "real-time translation in audio and text as the callers are on the line", a company spokesman told AFP, adding that it has not been determined how many languages will be supported.

The translation will be enabled even if a call partner uses a non-Samsung smartphone, as the new model will use "on-device AI technology".

Speaking to someone in a foreign language using the feature will be "as simple as turning on closed captions" on streaming shows, Samsung said in a press release, with private conversations safely locked in on the phones.

Analysts said it would be a "significant achievement" and sign of real technological progress if Samsung's AI could offer real-time translation of voice calls -- but questioned how it would work.

"The whole process will take at least three to four seconds during which the AI will understand an original content and translate into a foreign language and then verbalizes it to a listener," said Lee Won-kang, head of AI-based translation startup XL8's South Korea operation.

"We will have to actually see and use it ourselves to see how good it is, including whether translated contents will be spoken in AI voice or original voice."

AI sprawl

The unveiling comes as the South Korean firm is racing to develop its generative AI model called Samsung Gauss, which comes in three fields of language, code and image, and is currently used among company staff.

The company will incorporate the AI system into its wide range of products in the near future, it said, without specifying timing or the models.

Samsung's language AI will be able to help with everything from composing emails to summarizing documents, while the code AI will help developers to "code easily and quickly", the company said.

Samsung's AI on images will be capable of converting low-resolution images to high-resolution, and also both generating and editing images, it added.

Embedded generative AI "will change how we think about our phones forever", Choi Won-joon, head of Samsung's smartphone R&D sector, said in a statement.

While AI firms have hailed the technology's potential to facilitate major breakthroughs in science, medicine and public services, governments and watchdogs have raised alarms about data privacy and disinformation.

US President Joe Biden issued an executive order last month on regulating AI, the latest in a series of government measures aimed at curtailing the potential dangers of the technology.

Samsung said in a statement on Thursday that GalaxyAI will run on its devices, and "private conversations never leave your phone".



OpenAI, Anthropic Sign Deals with US Govt for AI Research and Testing

OpenAI logo is seen in this illustration taken May 20, 2024. (Reuters)
OpenAI logo is seen in this illustration taken May 20, 2024. (Reuters)
TT

OpenAI, Anthropic Sign Deals with US Govt for AI Research and Testing

OpenAI logo is seen in this illustration taken May 20, 2024. (Reuters)
OpenAI logo is seen in this illustration taken May 20, 2024. (Reuters)

AI startups OpenAI and Anthropic have signed deals with the United States government for research, testing and evaluation of their artificial intelligence models, the US Artificial Intelligence Safety Institute said on Thursday.

The first-of-their-kind agreements come at a time when the companies are facing regulatory scrutiny over safe and ethical use of AI technologies.

California legislators are set to vote on a bill as soon as this week to broadly regulate how AI is developed and deployed in the state.

Under the deals, the US AI Safety Institute will have access to major new models from both OpenAI and Anthropic prior to and following their public release.

The agreements will also enable collaborative research to evaluate capabilities of the AI models and risks associated with them, Reuters reported.

"We believe the institute has a critical role to play in defining US leadership in responsibly developing artificial intelligence and hope that our work together offers a framework that the rest of the world can build on," said Jason Kwon, chief strategy officer at ChatGPT maker OpenAI.

Anthropic, which is backed by Amazon and Alphabet , did not immediately respond to a Reuters request for comment.

"These agreements are just the start, but they are an important milestone as we work to help responsibly steward the future of AI," said Elizabeth Kelly, director of the US AI Safety Institute.

The institute, a part of the US commerce department's National Institute of Standards and Technology (NIST), will also collaborate with the U.K. AI Safety Institute and provide feedback to the companies on potential safety improvements.

The US AI Safety Institute was launched last year as part of an executive order by President Joe Biden's administration to evaluate known and emerging risks of artificial intelligence models.